AN ARCHITECTURE FOR A FUTURE WIRELESS ATM NETWORK

AN ARCHITECTURE FOR A FUTURE WIRELESS ATM NETWORK

a dissertation submitted to the department of electrical engineering and the committee on graduate studies of stanford university in partial fulfillment of the requirements for the degree of doctor of philosophy

By Bora Aydin Akyol June 1997

c Copyright 1997 by Bora Aydin Akyol All Rights Reserved

ii

I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Donald C. Cox (Principal Adviser)

I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Nick McKeown

I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Mary G. Baker

Approved for the University Committee on Graduate Studies:

iii

Abstract Communicating without being attached to a tether is appealing to many subscribers of wireless communication services that exist today. This dissertation describes an architecture for a Wireless Asynchronous Transfer Mode (ATM) network that will expand the range of services and the amount of resources available to wireless users in the future. Existing ATM networks are designed to support wireline users with xed locations; consequently, current ATM protocols do not implement Registration, Hando, Connection Setup and Rerouting functions that are required to support wireless users. Registration and Connection Setup are required to locate a user during information delivery. Hando provides true mobility to wireless users and allows them to move beyond the coverage of a single wireless access point. Rerouting is required to maintain connectivity to the network during a hando event. Overlay and Migratory Signaling are developed in this dissertation to implement registration, connection setup and hando functions in an ATM network context to support mobility of wireless users. Overlay Signaling uses switched ATM connections to encapsulate mobility related signaling messages between wireless-aware interworking nodes at the edges of the backbone ATM network and does not require any changes to the existing ATM protocols. Overlay Signaling utilizes a cell forwarding based approach to reroute user connections during a hando. Cell forwarding allows Overlay Signaling to maintain compatibility with the existing ATM protocols. Migratory Signaling de nes a new ATM signaling protocol that remains backward compatible with existing ATM protocols while supporting wireless users. Migratory Signaling uses the Nearest Common Node Rerouting (NCNR) algorithm iv

to reroute user connections during a hando. NCNR is a new rerouting algorithm developed in this dissertation and is based on the partial re-establishment of existing user connections. Our wireless ATM architecture provides wireless access to a backbone ATM network. In order to be compatible with the backbone ATM network, the wireless access points need to support multiple trac types with dierent priorities and quality of service requirements. Dynamic Resource Allocating Multiple Access (DRAMA), developed in this dissertation, is a medium access control and resource allocation protocol that supports multiple users, multiple connections per user and service priorities and is fully compatible with existing ATM protocols.

v

Acknowledgements I would like to express my deepest gratitude to my research advisor, Professor Donald C. Cox for his valuable support and guidance. I thank my associate advisor Professor Dale Harris for his support throughout my years at Stanford. I wish to thank Professors Peter Glynn, Jennifer Widom for participating in my orals committee. I am grateful to Professor Mary Baker for agreeing to be the chairperson in my oral dissertation defense and also for being a member of my reading committee. I thank Professor Nick McKeown for being a member of my reading committee. I thank Vern Paxson of LBL, Murad Taqqu of Boston University, Micheal Devetsikiotis, John Lambadaris of Carleton University and Changcheng Huang of Nortel for their comments on the self similar trac models in Chapter 5. My gratitude goes to C-K Toh of Cambridge University on his comments about Nearest Common Node Rerouting discussed in Chapter 4. I enjoyed the discussions with my colleagues in the \wireless" group: Sung Chun, Byong-Jo Kim, Dae-Young Kim, Derek Lam, T. Andy Lee, Yumin Lee, Tim Schmidl, Je Stribling, Daniel Wong and P. Bill Wong. Without my family, I would not be here. My father, may he rest in peace, has always been an inspiration to me, I hope he is watching. My mother and stepfather have always encouraged and supported me. Finally, my greatest thanks go to my wife, Noelle, who has always encouraged me and our son, Aydin, who has provided more than enough motivation for my work. Without their love and support, my work would have been much more dicult. It is to Noelle and Aydin that I dedicate this dissertation. vi

Contents Abstract

iv

Acknowledgements

vi

1 Introduction

1

1.1 Challenges in the Wireless ATM Network . . . . . . . . . . . 1.1.1 Challenges Related to the Mobility of Wireless Users 1.1.2 Providing Access to the Wireless ATM Network . . . 1.2 Outline of the Dissertation . . . . . . . . . . . . . . . . . . . 1.3 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

2 Wireless ATM Network Architecture

2 3 4 5 6

8

2.1 De nition of a Zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.2 Components of the Wireless ATM Network . . . . . . . . . . . . . . . 12 2.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3 Signaling in the Wireless ATM Network

3.1 The Overlay Approach to Wireless ATM Signaling . . . . . . . . . . . 3.1.1 ATM Connection Setup Procedure . . . . . . . . . . . . . . . 3.1.2 Registration Using Overlay Signaling . . . . . . . . . . . . . . 3.1.3 Call Setup Using Overlay Signaling . . . . . . . . . . . . . . . 3.1.4 Hando using Overlay Signaling . . . . . . . . . . . . . . . . . 3.1.5 Supporting Wireline and Wireless User Interaction in Overlay Signaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

14 15 17 18 21 23

30

3.2 The Migratory Approach to Wireless ATM Signaling . . . . . . . . . 3.2.1 Registration using Migratory Signaling . . . . . . . . . . . . . 3.2.2 Inter-zone Hando using the Migratory Signaling . . . . . . . 3.2.3 Connection Setup using Migratory Signaling . . . . . . . . . . 3.3 Estimated Signaling Overhead Required to Support Wireless Users . 3.3.1 Performance of the Overlay Signaling Approach . . . . . . . . 3.3.2 Performance of the Migratory Signaling Approach . . . . . . . 3.3.3 Comparison of the Overlay and Migratory Signaling Methods 3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4 Rerouting in the Wireless ATM Network

4.1 The Hando Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Rerouting for Inter-zone Hando . . . . . . . . . . . . . . . . . . . . 4.2.1 Rerouting for Overlay Signaling . . . . . . . . . . . . . . . . . 4.2.2 Nearest Common Node Rerouting (NCNR) for Migratory Signaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Preserving the Cell Sequence in Rerouting . . . . . . . . . . . . . . . 4.3.1 Preserving the Cell Sequence using Overlay Signaling . . . . . 4.3.2 Preserving the Cell Sequence using NCNR and Migratory Signaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 A Comparison of Existing Rerouting Algorithms for Wireless ATM Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Description of Alternate Rerouting Approaches . . . . . . . . 4.4.2 Comparison of Rerouting Algorithms . . . . . . . . . . . . . . 4.4.3 Quantitative Comparison of Rerouting Algorithms . . . . . . . 4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5 Medium Access in the Wireless ATM Network

5.1 Wireless ATM Network Environment . . . . . . . . . . . 5.2 Dynamic Resource Allocating Multiple Access Algorithm 5.2.1 Medium Access Control in DRAMA . . . . . . . . 5.2.2 Resource Allocation in DRAMA . . . . . . . . . . viii

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

31 32 34 39 46 48 50 52 55

57 59 60 61

65 72 72 74 75 76 78 84 87

88 89 91 91 93

5.3 Performance Estimation of the DRAMA Algorithm . . . . . . . . . . 96 5.3.1 Simulation Environment . . . . . . . . . . . . . . . . . . . . . 96 5.3.2 Simulation Parameters . . . . . . . . . . . . . . . . . . . . . . 99 5.3.3 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . 103 5.4 A Comparison of Medium Access Control and Resource Allocation Algorithms for Wireless ATM Networks . . . . . . . . . . . . . . . . . 110 5.4.1 Multiple-Services Dynamic Reservation (MDR) versus DRAMA111 5.4.2 Polling Multiple Access versus DRAMA . . . . . . . . . . . . 113 5.4.3 DQRUMA versus DRAMA . . . . . . . . . . . . . . . . . . . 114 5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

6 Conclusion

154

A ATM Signaling Message Lengths

158

B Derivation of Equations 3.1 through 3.3

159

C Migratory Signaling Messages

161

6.1 Summary of Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 154 6.2 Topics for Future Research . . . . . . . . . . . . . . . . . . . . . . . . 156

C.1 REGISTER . . . . . . . . . . . . C.2 REGISTER COMPLETE . . . . C.3 RECORD DELETE . . . . . . . C.4 REGISTER DENY . . . . . . . . C.5 HANDOFF INIT . . . . . . . . . C.6 HANDOFF INFO . . . . . . . . . C.7 HANDOFF CHANNEL ASSIGN C.8 HANDOFF REROUTE . . . . . C.9 REROUTE CONNECTION . . . C.10 REROUTE COMPLETE . . . . C.11 HANDOFF COMPLETE . . . . C.12 HANDOFF FAIL . . . . . . . . . ix

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. 161 . 161 . 162 . 162 . 162 . 162 . 164 . 164 . 164 . 164 . 166 . 166

C.13 RECORD UPDATE . . . . C.14 LOCATION REQUEST . . C.15 LOCATION REPLY . . . . C.16 REQUEST DENIED . . . . C.17 MOBILE SETUP . . . . . . C.18 MOBILE SETUP DENIED C.19 MOBILE ACCEPT . . . . . C.20 MOBILE START . . . . . . C.21 MOBILE CLEAR . . . . . . C.22 MOBILE CLEAR ACK . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

Bibliography

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. 166 . 166 . 168 . 168 . 168 . 171 . 171 . 171 . 171 . 171

173

x

Chapter 1 Introduction Communication is a process by which information is exchanged between individuals through a common system of symbols, signs, or behavior. Telecommunication is communication at a distance. Within the last decade, the world of telecommunications has started to change at a rapid pace. Data trac, where the information is transmitted in the form of packets and the ow of information is bursty rather than isochronous, now accounts for almost 250 Mbits/sec of the trac that is transmitted over the backbone telecommunication networks[1]. In addition to data trac, video trac made possible by low cost video digitizing equipment such as the QuickCam([2]) is also on the rise. An additional contributing factor to the change in telecommunications is the almost unlimited bandwidth provided by modern ber optical transmission equipment. Asynchronous Transfer Mode (ATM) technology is proposed by the telecommunications industry to accommodate multiple trac types in a very high speed wireline network. Brie y, ATM is based on very fast (on the order of 2.5 Gbits/sec or higher[3]) packet switching technology with 53 byte long packets called cells being transmitted through wireline networks running perhaps on ber optical equipment. Due to the xed packet size and the very fast packet switching, ATM meets very strict timing and delay requirements. This makes the transmission of time-sensitive trac, such as voice, through the ATM network possible. Since ATM is based on packet switching, it also accommodates data trac. ATM networks are designed to 1

CHAPTER 1. INTRODUCTION

2

support multiple trac types with dierent priorities and quality of service requirements. They are rst expected to be deployed in the backbone networks and then progress to the edge of the current telecommunications networks. Wireless telecommunications networks have broken the tether in the wireline networks and allow users to be mobile and still maintain connectivity to their of ces, homes, etc[4]. The wireless networks are growing at a very rapid pace; Personal Handiphone Service (PHS) in Japan has taken on ve million subscribers by the end of 1996. The GSM based cellular phones are being deployed in Europe, Asia, Australia and North America[5]. Wireless network subscribers demand not only voice but also data connectivity to the wireline networks such as the Internet. Existing digital cellular networks cater mainly to voice subscribers and oer only limited data capabilities including some short messaging and data rates on the order 14.4 Kbits/sec. Higher bit rate wireless technologies for data transmission have not gained wide-spread acceptance due to availability and connectivity limitations. A Wireless ATM Network provides a natural wireless counterpart to the development of ATM based wireline networks by providing full support for multiple trac types including voice and data trac in a wireless environment. This dissertation explores a possible architecture that may be used to implement a wireless ATM network. Supporting wireless users in an ATM network presents a unique set of challenges to the existing ATM protocols. These challenges will be discussed in the next section. In the following sections, we summarize the contents of each chapter and our contributions.

1.1 Challenges in the Wireless ATM Network Our ATM network architecture is designed to support wireless users. Supporting wireless users presents two sets of challenges to the ATM network. The rst set includes problems that arise due to the mobility of the wireless users. The second set is related to providing access to the wireless ATM network.


3

1.1.1 Challenges Related to the Mobility of Wireless Users The ATM standards proposed by the International Telecommunications Union (ITU) are designed to support wireline users at xed locations[6]; on the other hand, wireless users are mobile. Current ATM standards do not provide any provisions for support of location lookup and registration transactions that are required by mobile users[6]. They also do not support hando and rerouting functions that are required to remain connected to the backbone ATM network during a move. The user identi cation numbers in the wireline networks may be used for routing of connections to the user; in contrast, the identi cation number for a wireless user may only be used as a key to retrieve the current location information for that user. The location information for wireless users is usually stored in a database structure that is distributed across the network[7, 8, 9]. This database is updated by Registration transactions that occur as wireless users move within the wireless network. During a Connection Setup, the network database is used to locate and route connections to the user. If a wireless user moves while (s)he is communicating with another user or a server in the network, the network may need to transfer the radio link of the user between radio access points in order to provide seamless connectivity to the user. The transfer of a user's radio link is referred to as Hando. During a hando event, the user's existing connection may need to be rerouted in order to meet delay, quality of service or cost criteria or simply to maintain connectivity between two users or a server and wireless users. Since the existing ATM protocols are designed for wireline networks with xed users, support for Rerouting of existing user connections is not included in the ATM standards. Rerouting is critical to wireless networks which need to maintain connectivity to a wireless user through multiple, geographically dispersed radio access points. In this dissertation we describe the Overlay and Migratory Signaling protocols that are designed to implement mobility related functionality in an ATM network. We also describe Nearest Common Node Rerouting for rerouting connections in the wireless ATM network.


4

1.1.2 Providing Access to the Wireless ATM Network A key bene t of a wireless network is providing tetherless access to the subscribers. The most common method for providing tetherless access to a network is through the use of radio frequencies. There are two problems that need to be addressed while providing access to an ATM network by means of radio frequencies:

Error Performance of the Radio Link: ATM networks are designed to uti-

lize highly reliable ber optical or very reliable copper based physical media. These physical links utilize digital transmission techniques where information is encoded into bits. The probability of bit errors in modern wireline networks are on the order of 10? ; hence, ATM does not include error correction or checking for the user information portion of an ATM packet. Compared to the wireline networks, wireless networks may achieve average bit error rates on the order of 10? to 10? . In order to support ATM trac in a wireless ATM network, the quality of radio links needs to be improved through the use of equalization, diversity and error correction and detection to a level that is closer to the wireline networks. There are a number of solutions that combine these techniques to improve the error performance of wireless networks. Some of these solutions may be found in [4, 10, 11, 12, 13]. The choice of the physical radio transmission technology for the wireless ATM network is beyond the scope of this dissertation and will not be considered further. 9

3

6

Medium Access for Wireless ATM Networks: A wireless ATM network needs

to support multiple trac types with dierent priorities and quality of service guarantees. In contrast to the ber optical media in the wireline networks, radio bandwidth is a very precious resource for the wireless ATM network. A medium access control protocol that supports multiple users, multiple connections per user and service priorities with quality of service requirements must be developed in order to maintain full compatibility with the existing ATM protocols. This medium access protocol needs to make maximum use of the shared radio resource and needs to achieve full utilization of the radio frequencies in a variety of environments.


5

We propose the Dynamic Resource Allocating Multiple Access protocol in this dissertation for medium access control and resource allocation in wireless ATM networks.

1.2 Outline of the Dissertation This dissertation proposes a wireless ATM network architecture for a next generation wireless network that is designed to interface with the wireline networks based on ATM technology. We focus on problems related to the support of mobility in the wireless ATM network and problems related to controlling access to the shared radio resources in the wireless network. Chapter 2 introduces our wireless ATM network architecture and the terminology used in this dissertation. It introduces the zone concept and de nes basic components of the wireless ATM network. Chapter 3 describes two alternatives for implementing signaling in the wireless ATM network to support wireless users. Overlay Signaling is designed to support mobility using the existing ATM protocols. It implements this support using a signaling network overlaid on the backbone ATM network using the Zone concept discussed in Chapter 2. Overlay Signaling is the only wireless ATM network signaling protocol proposed in the literature that is able to function using the current ATM signaling protocols. On the other hand, Migratory Signaling updates the ITU Q.2931 signaling protocol ([6]) proposed for ATM networks to include built-in support for wireless users. When compared to Overlay Signaling, Migratory Signaling uses less bandwidth for signaling and does not need additional signaling circuits that are dedicated for support of wireless users. Overlay Signaling implements the support for wireline and wireless user interaction by using service gateways. Migratory Signaling provides built-in support for wireless users. Chapter 4 focuses on the problem of dynamic rerouting of user connections during a hando event. We present a rerouting method based on cell forwarding for wireless ATM networks that use Overlay Signaling. Also in Chapter 4, we present the Nearest Common Node Rerouting (NCNR) protocol for dynamic rerouting of user connections in wireless ATM networks that use Migratory Signaling. NCNR is


6

based on rerouting a user connection at the node that is the closest common ancestor of the zones that are involved in the hando. NCNR makes maximum re-use of the existing user connection during rerouting and saves network resources. When compared to alternate rerouting schemes previously proposed in the literature, NCNR consistently involves fewer network nodes and requires fewer signaling messages for rerouting. Also in Chapter 4, we identify the dierent needs of time sensitive and throughput dependent trac types during a hando and show how hando may be implemented to better suit the needs of these types of trac. We propose Dynamic Resource Allocating Multiple Access (DRAMA) as a medium access control and resource allocation protocol for wireless ATM networks in Chapter 5. DRAMA accommodates multiple trac types, multiple connections per user with dierent quality of service criteria. An analysis of the protocol together with simulation results is also presented in this chapter. According to our simulations, DRAMA achieves resource allocation success rates of over 90 percent under a variety of operating conditions. The results of our simulations indicate that DRAMA performs better than or equal to previously proposed medium access control protocols for wireless ATM networks while providing the added bene ts stated in Chapter 5. Also in this chapter, we develop baseline trac models for wireless ATM networks and simulation techniques that are suitable for simulating self similar trac patterns. Finally, Chapter 6 summarizes the main results of this dissertation and suggests areas for future research.

1.3 Contributions Here is a brief summary of contributions made in this dissertation.

Overlay and Migratory Signaling Protocols were created to support mobile users in a Wireless ATM Network (Chapter 3)

Nearest Common Node Rerouting was designed to support dynamic rerouting of user connections in the wireless ATM network (Chapter 4).


7

Dynamic Resource Allocating Multiple Access was designed to perform ecient and fast medium access control of the shared radio resources (Chapter 5).

Quantitative and qualitative analyses of all proposed protocols and algorithms were performed and comparisons with alternate schemes when available were presented.

Chapter 2 Wireless ATM Network Architecture This chapter introduces our wireless ATM network architecture. Section 2.1 describes the zone concept. In Section 2.2, the components of the wireless ATM network and the functions of these components are described. Section 2.3 provides a summary of the chapter.

2.1 De nition of a Zone Our wireless ATM network architecture is based on the Zone concept. An example of a zone in the wireless ATM network is depicted in Figure 2.1. A zone consists of radio ports, radio port controller(s), wireless to ATM interworking equipment, possibly a database and the physical links that interconnect the parts of the zone. The software that manages the physical equipment in the zone and maintains connectivity to the backbone ATM network is referred to as the Zone Manager. The wireless ATM network is designed as a micro-cellular network for the reasons described in [10, 14]. The typical coverage of a radio port in a micro-cellular network varies between half a mile to one mile[10]; therefore, a fairly large number of radio ports are required in order to maintain full coverage of a given geographical area. Consequently, the radio ports in a micro-cellular network must be economical radio 8

CHAPTER 2. WIRELESS ATM NETWORK ARCHITECTURE

Radio Port Radio Port Radio Port Radio Port

Radio Port

Radio Port Controller To the ATM Network

Radio Port

Radio Port Controller Wireless-to-ATM Network Interface Database

Figure 2.1: A Zone in the Wireless ATM Network

9


10

modems that are small enough to be placed on rooftops and utility poles[4, 10]. The radio ports are responsible for:

Maintaining physical connectivity to the wireless users and performing radio access point functionality,

Transmitting information from the network to the wireless users and from the wireless users to the wireless ATM network,

Modulation and demodulation of radio signals into Wireless ATM radio packets.

The radio port controllers control access to the shared radio resources of the wireless ATM network and assist the zone manager during a hando event. Radio port controllers are responsible for allocation of resources to wireless users and relaying signaling requests from the wireless users to the zone manager. The radio port controllers may optionally perform segmentation and re-assembly of ATM cells into wireless ATM radio packets. A wireless ATM radio packet may contain from a fraction of an ATM cell to multiple ATM cells depending on the structure of the wireless ATM radio link. It also includes error-correction and detection overhead that may be required to assure the integrity of the user information. In a wireless ATM network where users are globally mobile, the tracking of users is one of the major functions of the wireless network. Each zone may have a database that is used to support the tracking process[9, 15]. This database is partitioned into two segments. One segment is reserved for the users that are permanently registered in that zone, i.e. the \home" segment. The second segment is for the users that are visiting the zone. The wireless ATM network databases do not keep track of the radio ports that connect the users to the network. The radio port controller keeps track of radio port and active user information. The user identi cation numbers determine the location of user's permanent registry or home database. The nearest database is the rst place to be searched in attempting to locate user information. If the desired information is not found at the nearest database then the permanent home database for that user is queried.


11

We will refer to this architecture as the \two-tier database architecture." The European GSM and U.S. IS-41 digital cellular standards of \Home Location Registers/Visiting Location Registers" (HLR/VLR) are examples of two-tier database architectures[5, 15]. One important advantage of a two-tier architecture as described above is the two-step deterministic search where a user's pro le is retrieved in at most two database look-ups. An alternative to a two-tier database architecture is a hierarchical structure which takes advantage of the locality and motion patterns of users and organizes the distribution of data accordingly [7, 8, 9]. In this dissertation, we assume that the wireless ATM network uses a two-tier (HLR/VLR) database architecture. The zone manager software runs on the Wireless to ATM interworking equipment and resembles the operating system of a personal computer. The zone manager is responsible for:

Handling the signaling transactions that are necessary to support wireless users in the wireless ATM network,

Performing routing of user connections between the radio port controllers and the rest of the wireless ATM network,

Controlling the radio port controllers and managing the connections between the radio port controllers,

Mediating access to the backbone network and for signaling between the wireless users and the rest of the wireless ATM network.

If the segmentation and re-assembly of ATM cells into wireless ATM radio packets are not being performed by the radio port controller, the wireless to ATM interworking equipment must perform segmentation and re-assembly of ATM cells into wireless ATM radio packets as described above.


12

2.2 Components of the Wireless ATM Network Our wireless ATM network architecture consists of zones, wireless ATM network backbone and gateways to the wireline ATM network(s) as depicted in Figure 2.2. The zones of the wireless ATM network are responsible for supporting wireless users. Network Database ATM Network Gateway

Wireline ATM Network

Network Operations & Control

Wireless ATM Network Backbone

Zone 1

Zone 2

Zone 3

Figure 2.2: Components of the Wireless ATM Network Each zone incorporates the signaling functionality required to support mobile users. Via the use of zones, our wireless ATM network architecture is a completely distributed network. By dividing the wireless ATM network into zones, we also reduce the addressing granularity of the wireless ATM network. The radio ports and radio port controllers have only local signi cance within the zone. In terms of locating and routing connections to wireless users, the wireless ATM network only considers the zone of the user and not the particular radio port. In the other direction, the


13

location of the user needs to be updated only when the user moves between the zones which signi cantly cuts on the signaling trac as described in Chapter 3. The network operations and control unit given in Figure 2.2 is responsible for monitoring the performance of the wireless ATM network and does not perform any signaling related functions. The ATM network gateway manages the ow of information to and from the wireless ATM network to the wireline ATM networks. The ATM network gateway is necessary to support connections between the wireline ATM network users and wireless users and is responsible for performing location resolution functionality for wireline network users as described in [8].

2.3 Summary In this chapter, we presented our wireless ATM network architecture. The building blocks of the network were identi ed, illustrated and de ned. We brie y mentioned the database aspects related to the signaling in the wireless ATM network. We refer the reader to [8, 9] for excellent discussions of database issues in wireless telecommunications networks.

Chapter 3 Signaling in the Wireless ATM Network In this chapter, we de ne how signaling support for registration, connection setup and hando transactions can be implemented in a wireless ATM network. Brie y, Registration is required to locate a user during information delivery. Connection Setup is used to establish connections to other users or servers in the wireless network. Hando provides true mobility to wireless users and allows them to move beyond the coverage of a single wireless access point. Existing ATM signaling protocols do not support Registration,Connection Setup and Hando transactions that are required to support wireless users[6]. In order to support wireless users in our ATM architecture, we need to adapt the registration, connection setup and hando procedures used in existing wireless communication networks([5, 15]) to function in an ATM network. There are two possible approaches for implementing support of mobility in an ATM network. The rst approach, called the \Overlay Signaling" approach, uses ATM connections to transport mobility-related signaling messages between the zones in the wireless ATM network and does not require any changes to the existing ATM protocols. The resulting signaling network is then overlaid on top of the existing ATM network. Our motivation for implementing an overlay signaling network is to 14

CHAPTER 3. SIGNALING IN THE WIRELESS ATM NETWORK

15

remain compatible with the existing ATM protocols. Since there are no modi cations to the ATM protocols, the overlay signaling approach does not require any modi cations to existing ATM infrastructure. Our second approach is called the \Migratory Signaling" approach. Migratory Signaling modi es existing ATM signaling protocols to accommodate wireless users together with wireline users without the need for an overlay signaling network. Migratory Signaling aims to minimize the overhead incurred in supporting mobile users due to the overlay signaling. These two approaches may be used together as a migratory path in evolving to a single global wireless ATM network. We describe the overlay signaling approach in the following section and the migratory signaling approach in Section 3.2. We conclude by presenting qualitative and quantitative comparisons of overlay and migratory signaling approaches in Section 3.3.

3.1 The Overlay Approach to Wireless ATM Signaling In this section we discuss the network transactions that are related to supporting user mobility in a wireless ATM network and suggest how they can be implemented using the ATM User Network Interface (UNI) signaling protocol[16]. The network transactions related to user mobility are: 1

Registration, Call Setup, Hando. ATM UNI (version 3.1) is an implementation agreement proposed by the ATM Forum and coincides with the ITU-Q.2931 signaling protocol implementation. In this text the UNI signaling messages will be typeset in BOLDFACE characters. Note also that the zones of the wireless network will appear as \users" to the xed ATM network hence use of ATM UNI signaling is justi ed. 1


16

There are two networks and two kinds of signaling in the Overlay wireless ATM signaling network architecture. The zones are interconnected by an ATM network. The zone managers must use the ATM signaling in order to establish connections through the ATM network. The radio ports and portables belong to the wireless network, and the zone managers exchange wireless network signaling messages in order to support the needs of the users and the wireless ATM network. The wireless netRadio Ports

Radio Ports * Radio Port Controllers

Radio Port Controllers Zone Manager (ZONES) ATM Network

Wireless Network Signaling Messages Using ATM ITU Q.2931B Signaling Messages Using ATM

Backbone

Zone Manager (ZONES) ATM Network

Backbone

Physical Link

Wireless Network Layers

Physical Link Layer Protocol

Physical Link

(*) The radio ports and port controllers can only communicate with the corresponding zone managers. The communication between radio ports and port controllers is unspecified.

Figure 3.1: Protocol Layers in the Wireless and ATM Networks. The radio ports and the port controllers may only communicate with the zone managers. The links between the radio ports and the port controllers are not speci ed. work signaling messages are encapsulated into ATM cells and transmitted through the ATM network to their destinations. See Figure 3.1 for a depiction of connection types and protocols between the layers in the wireless and ATM networks. A procedure that is common to all of the wireless network transactions is establishing a connection through the xed ATM network. The current ATM signaling


17

speci cations describe how an ATM connection is established in a xed network context[6, 16] using current ATM protocols between two ATM network entities. In this chapter, registration, call setup and hando will be implemented using the ATM Connection Setup Procedure without modi cation in an overlay manner. For the sake of completeness we describe the ATM Connection Setup Procedure in the next section.

3.1.1 ATM Connection Setup Procedure An ATM connection in a xed network context is established between two parties directly connected to the ATM network, the calling party that initiates the connection and the called party that is the destination. The calling/called parties will be zone managers in the wireless ATM network context. If the radio ports, and the radio port controllers are also interconnected by the ATM network, then the same procedure could be used by the radio ports and radio port controllers. The ATM Connection Setup Procedure is discussed in the following paragraphs: 1. SETUP: The setup message is sent from the calling party to the ATM network and from the ATM network to the called party. This message includes the called party address information, ATM user cell rate (ATM trac descriptor), broadband bearer capability and quality of service (QoS) parameters. The ATM network is responsible for selecting a connection identi er and sending a setup message to the called party. For network management purposes the connection is also assigned a call reference by the ATM network. This call reference is used in all of the following messages to refer to the ongoing connection setup. 2. CALL PROCEEDING: The ATM network upon receiving the SETUP message will determine whether access to the requested service is authorized. If the access is authorized then the network sends a CALL PROCEEDING message to the calling user. The calling user then waits for the connection.


18

3. CONNECT: The connect message is sent by the called party to the ATM network to indicate call acceptance. The ATM network then sends a similar message to the calling party to indicate that a connection is established. If the called party is not able to send a CONNECT message in the allocated time, it sends a CALL PROCEEDING message to the ATM network while processing the connection request. If a CALL PROCEEDING message is sent by the called user to the ATM network, the ATM network waits for a CONNECT message until the connection setup timer expires. 4. CONNECT ACKNOWLEDGE: The calling user sends this message to the ATM network to acknowledge the successful connection, and goes into active state. The ATM network sends this message to the called party to acknowledge the successful connection setup. The state diagram for ATM connection setup is given in Figure 3.2. The ow of ATM signaling messages for a successful ATM connection setup is given in Figure 3.3. In a network where multiple hops are needed to establish a connection, this procedure is repeated at every hop. If an intermediate node does not have the resources to establish the connection, a RELEASE COMPLETE message will be sent to the initiating party to clear the call. If the called user is busy or if the incoming request is rejected the called user will send a RELEASE COMPLETE message which will include the reason for refusal. An active connection is terminated by the user by sending a RELEASE message to the network and the network will respond with a RELEASE COMPLETE message and both parties will enter the idle state.

3.1.2 Registration Using Overlay Signaling In the following sections we show how the ATM Connection Setup procedure is used to implement Registration, Call Setup, and Hando transactions for the wireless network. Registration is performed to maintain information about the wireless user locations. The registration is performed as follows:

CHAPTER 3. SIGNALING IN THE WIRELESS ATM NETWORK Receive (Rcv) Call Proc. from Network

Send SETUP to Network

IDLE

CALL INIT

CALL PROC.

19

Rcv. CONNECT from Network Send CONNECT ACK.

ACTIVE

Time Expired Call Rejected

Call completed

Figure 3.2: ATM Connection SETUP state diagram 1. The registration process starts with the transmission of the user identi cation number (UID) and user's previous zone identi cation from the portable that enters a new zone. The portable recognizes the zone change by comparing the current zone identi cation with the stored last zone identi cation. The radio ports are assumed to transmit zone identi cation beacon signals periodically to assist the registration process [10, 14]. After exchange of encryption keys between the portable and the radio port, user's password is also transmitted over the radio interface in an encrypted message. We also note that if public key cryptography is being utilized for encryption, the encryption public key may be sent in the clear. Discussion of encryption technologies is beyond the scope of this dissertation. 2. Upon receiving the UID and the authentication information, the zone manager (ZM) of the new zone establishes an ATM connection to the zone that contains the user pro le (\home" zone). The location of the user pro le may be obtained by querying the user's previous known zone. If the user pro le is

CHAPTER 3. SIGNALING IN THE WIRELESS ATM NETWORK Calling Party

20

ATM Network Called Party

SETUP SETUP CALL PROC. CONNECT

CONNECT

CONNECT ACK. CONNECT ACK.

Figure 3.3: Signaling Message Flow for ATM connection setup. stored in a hierarchical database structure then a search has to be performed in order to locate the record; however, the steps involved in each step of the search will the same as what will be described in the following paragraphs. The zone managers communicate by using the wireless network signaling protocol messages that are encapsulated into ATM cells. 3. After the ATM connection between the new and home zone managers is established, the zone manager of the user's current zone requests the user's authentication record and upon receipt of this record, the user is authenticated. 4. If authentication is successful, the user's pro le is updated with the new location information and the updated pro le is transferred to the current zone. 5. The user's pro le in the previous zone is deleted by establishing an ATM


21

connection to the previous zone manager. 6. If the authentication is unsuccessful then the user access is denied. 7. After the registration transaction is complete, the connection is released using RELEASE and RELEASE COMPLETE messages. Please see Figure 3.4 for message ows for a successful registration . 2

Portable

New Zone Home Zone Previous Zone Registration Request Establish ATM connection and obtain User Profile from Home Database

User Profile

Notify user If user is authenticated delete profile from previous zone.

Figure 3.4: Overlay Signaling Message Flow for Registration. See footnote 2.

3.1.3 Call Setup Using Overlay Signaling The call setup procedure is used to establish a connection between two wireless network users. In this section, the originating zone refers to the calling user's zone The release of Signaling Connections are not shown for the sake of clarity of gures. They are included in our calculations. 2


22

and the destination zone refers to the called user's zone. The call setup proceeds as follows: 1. Called user identi cation number (CUID) is transmitted from the portable to the originating zone manager together with call setup parameters such as required bandwidth, trac type, etc. 2. The originating zone manager (OZM) forms an ATM UNI protocol SETUP message using the incoming call parameters and proceeds as follows: (a) OZM requests the called user's location information from the home zone in the wireless ATM network. The procedure for this process is similar to what is described in Sections 2.1, 3.1.2 and gure 3.4. (b) When the called user's location information is retrieved from the home zone, an ATM connection is established to the called user's current zone manager using the ATM connection setup procedure. (c) After the connection is established, the originating zone manager (OZM) passes the calling user's identi cation number to the destination zone manager (DZM) and requests a connection setup. If the calling user's identi cation number is not in the call blocking list of the called user then the called user is paged, if the page is successful, the DZM requests an ATM connection between the two users. If the page is not successful then the called user is assumed inactive and the proper record is updated. If the calling user is in the call blocking list of the called user then connection is rejected. (d) Pending ATM user connection establishment, the calling user and the called user are noti ed and the connection is established. (e) If the called user is busy, then the DZM noti es the OZM of the \busy" state and call attempt is terminated. The connection between the two zone managers is released. 3. After the call is complete, the ATM connection is released using RELEASE and RELEASE COMPLETE ATM signaling messages.


23

See Figure 3.5 for the call setup transaction ow for a successful call. Remarks:

Call blocking may be checked at the same time as the user's pro le is obtained. The connection setup for the user connection may be initiated by the OZM. Zone Manager

Calling User

Call Setup Request

ATM Network

User Database

Zone Manager

Called User

ATM Conn. Setup to called user DB.

User Loc. Data

ATM Conn. Setup to called user zone.

Page User Accept Conn.

Respond

Setup User Conn.

Alert Calling User

Respond User Conn. Established

User Conn. Established

TWO WAY USER CONNECTION ESTABLISHED

Figure 3.5: Overlay Signaling Message Flow for Call Setup

3.1.4 Hando using Overlay Signaling Hando is the transfer of a user's radio link between radio ports in the network (Refer to Chapter 4 for details on hando). Hando may occur due to multiple reasons:


24

The user's radio link to a radio port may be failing due to movement and environmental factors.

The user's radio signal may be causing interference to other users in the vicinity.

The user may gain access to better resource availability by using an alternate

radio port. In our wireless ATM network architecture, two types of hando are possible. An intra-zone hando occurs between two radio ports in the same zone. This hando is handled by the zone manager and the radio port controllers in the zone. As described in Section 2.1, the zone is the basic building block of the wireless ATM network and the network does not keep track of radio ports or radio port controllers. Since the user remains in the same zone, the intra-zone hando does not generate external wireless ATM network signaling messages. The inter-zone hando occurs between two radio ports that belong to two dierent zones. In this type of hando, the network needs to perform rerouting and location updates to maintain the status of the user's connection(s) since the user changes its current zone. Due to the wireless ATM network involvement, the inter-zone hando generates external network signaling trac. In this chapter, we are primarily concerned with wireless ATM network signaling; therefore, we limit the discussion to inter-zone handos. The intra-zone hando is performed in a similar manner by the radio port controller without generating network signaling trac. The inter-zone hando process depends on its implementation in the wireless network. We will assume the following: Hando is portable initiated. The portables monitor the link quality in terms of received signal power to candidate radio ports and when the link to another port becomes better, that port is selected and hando is initiated[10, 14]. The link quality is determined by the portables because only the portables can determine the quality of links to multiple radio ports and decide on the best link. In contrast, a radio port can only monitor the link between itself and the portable.


25

The hando process may be initiated in two ways: The portable may tune into

the control channel of the candidate radio port and initiate a hando through the candidate port or the portable may use the existing link with the previous port to initiate the hando. The latter method is used more frequently in present systems[10]; however, both methods have merit and will be discussed.

Previous Link

Previous Radio Port

Radio Port Controller

Candidate Link

Portable

Candidate Radio Port

Radio Port Controller PREVIOUS ZONE

CANDIDATE ZONE

Figure 3.6: The Hando: Basic Picture Since we limit our discussion to the inter-zone handos, the candidate radio port and the previous radio port are in dierent zones. The candidate radio port is controlled by the \candidate zone manager." The previous radio port is controlled by the \previous zone manager." Please refer to gures 3.6, 3.7 and 3.8 for details. The hando will proceed as described below:

Hando through the Previous Port 1. In this case, the portable realizes that a link of better quality exists to a candidate radio port. The portable records the identity of the candidate port.


26

2. The portable sends a message to the previous zone manager (PZM) desiring a hando to the candidate radio port. The ATM address of the candidate zone is included in this message. Since we are considering inter-zone handos the candidate radio port is in a dierent zone. 3. The PZM establishes an ATM connection to the candidate zone manager (CZM) using the ATM connection setup procedure. 4. The PZM transfers a copy of the user pro le to the CZM, CZM assigns a channel to the user, relays the channel assignment information to the PZM. 5. The PZM contacts the end point for the user connection and requests rerouting to the candidate zone. Rerouting in the event of a hando is discussed in Chapter 4 of this dissertation. 6. Once the re-routing is complete, the PZM contacts the portable and relays the channel assignment information. 7. The portable tunes to the new channel and contacts the CZM. 8. The CZM and the portable verify the connection. After veri cation the CZM noti es the PZM of the successful hando. If the connection is not veri ed, the portable tunes to the previous channel and starts scanning for candidate ports. The CZM deallocates the assigned channel. An alternative to this procedure is to verify the radio channel and then perform the rerouting in order to save resources. However, since the hando is portable initiated and depends on power measurements to multiple radio ports, it is very likely to be successful[10]. 9. If the hando is successful, the PZM deletes user pro le.

Hando through the Candidate Port 1. The portable initiates a hando by establishing a link with the candidate radio port in the candidate zone and requesting a hando from the candidate zone


Portable

Previous Zone Mgr.

Hand-off

27

Candidate End point for Zone Mgr. user connection

Establish ATM Conn. Request Hand-off

Request Channel Assignment

Establish ATM Conn. Request Re-routing

Channel Assignment Data

Re-route complete

Verify Radio Connection Notify end point Connection Verified Re-routed data

Figure 3.7: Overlay Signaling Message Flow for Successful Hando, Case 1, Using Previous Port. See footnote 2. manager. This message includes the user's identi cation number and previous zone identi cation. 2. The candidate zone manager will acknowledge this message. The candidate zone manager (CZM) will establish an ATM connection to the previous zone manager (PZM) using the ATM connection setup procedure. The portable will tune back to its previous channel after the acknowledgment is received. 3. Following the connection establishment, the CZM will request a hando. The PZM will transfer the user pro le to the CZM using the ATM connection. The PZM shall also send a re-routing message to the ATM network so that the route to the candidate zone may be established. Rerouting of wireless ATM connections due to a hando event is discussed in detail in Chapter 4 of this dissertation.


28

4. The CZM shall provide the PZM with the channel assignment information for the portable. Pending re-routing success, the CZM will start to buer the information for the portable. If the re-routing is not successful the hando attempt is dropped, the PZM is noti ed. The PZM in turn noti es the portable. 5. If the re-routing is accomplished, the PZM sends the proceed with hando message to the portable that is still tuned to the channel in the previous zone. The proceed with hando message includes the channel assignment information for the candidate zone. If the portable cannot be raised at the previous zone, the CZM will be noti ed and the CZM will page the portable to establish a link. This page will also include the candidate channel assignment. 6. The portable, after receiving either the channel assignment information or the page from the CZM, will tune to the assigned channel and transmission is initiated. In either case the buered information will be passed on to the portable in the order received. If the accepted time limits for the connection are exceeded then some information such as expired voice packets may be dropped. 7. When the hando is completed and the stability of the new link is established, the ATM connection between the zone managers is released. Upon release of this connection the PZM will update the user's pro le using a procedure similar to registration (authentication is not needed in this case) and erase the information that is buered for the portable at the previous zone. The dierences between these two alternative methods are subtle; however, the former method (hando through the previous port) does not need synchronization to the control channel of the candidate port and uses the existing radio link to the previous port for hando related signaling. Note that if hando is being performed because the existing link is deteriorating, then the success of the former method is not assured; nevertheless, if the portables monitor the link quality frequently then this will be a minor problem. The latter method suers from diculty of carrying

CHAPTER 3. SIGNALING IN THE WIRELESS ATM NETWORK Portable 1

Alternate Zone Manager

Initiate

Previous Zone Manager

Hand-off Request

Hand-off Acknowledge

29

End Point for the User Connection

Re-routing User Information

Request

Acknowledge ATM Connection Setup

Acknowledge

User Data

User Data Sent to New Zone

Hand-off Proceed

Verify Radio Connection

Release Connection Release

Initiate Data Transfer

Hand-off Complete

Figure 3.8: Overlay Signaling Message Flow for Successful Hando, Case 2, Using Candidate Port. See footnote 2. on two connections at once but does not suer from the failing link problem of the latter. An interesting phenomenon that may occur during a hando happens when two wireless users that are connected to each other through the wireless network perform a hando at the same time. Even though this situation will be observed fairly rarely it still needs to be addressed. We will refer to this phenomenon as Simultaneous Hando. During a simultaneous hando, if at least one of the users stay within their zone (or perform an intra-zone hando), the network will perform the hando as described above. If both users perform an inter-zone hando, the previous zone managers (PZM) for both users will be responsible for forwarding the signaling messages related to the hando to the candidate zone managers (CZM). Note that in this case there already is a connection between the PZM and the CZM due to


30

the hando transaction and this does not bring any additional load to the network. Since the signaling messages are forwarded, no information is lost. The remaining parts of a simultaneous hando are handled appropriately using the procedures discussed above.

3.1.5 Supporting Wireline and Wireless User Interaction in Overlay Signaling In this section we discuss how the interaction between wireline and wireless users may be implemented in the Overlay Signaling approach. In the wireline networks, the user's identi cation number determines where the user is and how to route a call to the user. However, in a wireless environment, the user identi cation numbers do not yield any information about the user's location. In such an environment the user's number must be used to retrieve the location information. Note that if the wireless and ATM networks were separate entities in which no cross-interaction between users occurred, then the discussions in Section 3.1 would provide a way of implementing a self-contained wireless network supported by an ATM network. However, the users in the respective networks need to interact with each other, so the ATM signaling protocols need to be modi ed to understand the existence of mobile users. The ATM signaling protocols currently support only xed user numbers. These numbers are treated very much like a phone number to establish where to route a call to any given user. When wireline ATM users are allowed to interact with wireless users we face the following numbering system problem. When a wireline ATM user desires to connect to a user that is a wireless network subscriber, the called user's number will not yield any routing information to an ordinary ATM host. This implies that we need to have a numbering scheme that accommodates the wireless users. An interim solution is to de ne wireless user numbers as specially designated ATM numbers and have ATM to wireless user calls routed to designated wireless user switches that will re-route the calls to their nal destinations. A longer-term solution to this problem is given in [8]. The calls from wireless network subscribers


31

to wireline ATM hosts do not constitute a problem as the burden of establishing the connection lies with the wireless network which is aware of both kinds of numbering. This alternative is easy to implement and does not require any major modi cation to the ATM protocols. Because of ease of implementation this approach could provide a good interim solution. Unfortunately, this interim method will not scale well as the number of wireless network users grow and the wireless switching centers get congested. We might also face having to re-route calls across hundreds of miles to establish a connection to a user that is next door. This leads to our next topic which will relieve these problems. This completes the speci cation of the proposed implementation of a wireless ATM network that overlays the xed ATM network. In the next section we describe the Migratory Signaling Approach for wireless ATM network signaling.

3.2 The Migratory Approach to Wireless ATM Signaling Migratory Signaling is a protocol implementation for a wireless ATM network that satis es the goals listed below:

The signaling protocol shall support all the functions supported by the exist-

ing ATM signaling protocol and will require minimum change to the existing protocols.

The signaling protocol shall be migratory. It will support upgrading of the

network in phases or in regions while maintaining compatibility with the existing network through the use of service gateways as described in the previous section.

The migratory signaling approach implements a single signaling protocol for support of both wireless and wireline users. We focus on how Registration, Hando and Connection Setup are implemented in this new signaling protocol implementation in this section. We refer to the new protocol as the \Migratory Signaling Protocol" and


32

use the words user and terminal to refer to a wireless ATM network endpoint. We also assume that a wireless ATM user's identi cation number uniquely determines the permanent home database for that user as described previously (See Section 2.1). The support for wireline ATM users is de ned in ATM Forum and ITU documents [6, 16]. We therefore explain how the wireless users in the wireless ATM network are supported and how the interaction between the wireline and wireless users is implemented. The wireline ATM users are supported by incorporating the current ATM protocols into the Migratory Signaling Protocol.

3.2.1 Registration using Migratory Signaling Registration in the wireless ATM network is performed to keep the user location information current and is implemented as follows: 1. The user enters a new zone. The user terminal initiates a registration session with the zone manager for the new zone. After encryption key exchange as described in Section 3.1.2, the terminal sends the user's Identi cation Number (IDN) to the zone manager and goes to standby state. 2. The zone manager records the IDN and sends a REGISTER message to the user's permanent registration database . The REGISTER message contains user's identi cation number, the zone's ATM address, the user's permanent home database ATM address and other standard ATM signaling parameters. 3

3. The REGISTER message is transmitted through the ATM network using the ATM signaling virtual circuits. The permanent home database receives the message and authenticates the user. Upon authentication it updates its records and sends two messages: REGISTER COMPLETE to the user's new zone and RECORD DELETE to the user's previous zone to delete user's record . The REGISTER COMPLETE message contains the user IDN, the ATM 4

Throughout this section we will use BOLDFACE letters to dierentiate new wireless ATM signaling messages. See Appendix C for details on the contents of Migratory signaling messages. 4 RECORD DELETE message is not sent in the so-called lazy deregistration schemes. 3


33

address of the current zone, the user pro le and the ATM address of the home database of the user. The RECORD DELETE message contains the user IDN, the ATM address of the previous zone and the home database of the user. 4. If the authentication is not successful then the user's new zone is noti ed by sending a REGISTER DENY message. 5. If the registration is successful, the user's terminal is noti ed; for unsuccessful registration attempts the registration session is re-initiated. If the registration is unsuccessful a second time then the user is denied service and the network operations control is noti ed. The signaling message ow for registration is given in Figure 3.9. User

Current Zone Mgr.

Initiate Reg.

Auth. Key Exchange

Perm. Previous Home DB Zone Mgr.

REGISTER

REGISTER COMPLETE

RECORD DELETE

Figure 3.9: Migratory Signaling Message Flow for Successful Registration Note that there are no overlay ATM connections established for this process. All signaling messages are native wireless ATM signaling messages and use ATM signaling circuits to traverse the network.


34

3.2.2 Inter-zone Hando using the Migratory Signaling As in the discussion of hando given in Section 3.1.4, we concentrate on the interzone hando; therefore, the candidate radio port and the previous radio port are in dierent zones. We assume that the hando is portable-initiated and assisted, and that the hando may either be performed through the candidate or through the previous radio port. See [17, 18] and Chapter 4 for details on hando. The candidate radio port is controlled by the candidate zone manager (CZM). The previous radio port is controlled by the previous zone manager (PZM). Please refer to gures 3.6, 3.10 and 3.11 for details. The hando proceeds as described below:

Hando through the Previous Port 1. In this case, the portable realizes that a link of better quality exists to a candidate radio port. The portable records the identity of the candidate port. 2. The portable sends a message to the PZM requesting a hando to the candidate radio port. Since we are considering inter-zone handos the candidate radio port is in a dierent zone managed by the CZM. 3. The PZM sends a HANDOFF INIT message to the CZM. This message contains the ATM addresses for the PZM and CZM, and the user's IDN and pro le. 4. CZM receives the HANDOFF INIT message and assigns a channel to the user, relays the channel assignment information to the PZM by sending a HANDOFF CHANNEL ASSIGN message to the PZM. 5. The PZM noti es the end point for the user connection by sending a HANDOFF REROUTE message. The HANDOFF REROUTE message contains the address of the candidate zone and the user's IDN. 5

6. The end point for the user connection sends a REROUTE CONNECTION message to the CZM. The end point at this stage also stops sending information 5

The end point is de ned to be the terminating node for the user connection in the network.


35

to the PZM and starts buering the information. The rerouting is performed at the point where the route to the PZM and the route to the CZM forks. The CZM sends a REROUTE COMPLETE message to the PZM. See Chapter 4 for detailed description of rerouting due to a hando in wireless ATM Networks. 7. Once the re-routing is complete, the PZM contacts the portable and relays the channel assignment information. 8. The portable tunes to the new channel and contacts the CZM. 9. The CZM and the portable verify the connection. After veri cation the CZM noti es the PZM of the successful hando by sending a HANDOFF COMPLETE message. Upon receiving the HANDOFF COMPLETE, PZM releases the connection between itself and the end point. The CZM sends a HANDOFF COMPLETE to the end point to resume data transfer. If the connection is not veri ed, the portable tunes to the previous channel and starts scanning for candidate ports. The CZM deallocates the assigned channel and sends a HANDOFF FAIL message to the PZM and the end point. The end point starts sending the data to the PZM. Note that some time-sensitive information such as voice may be discarded if the hando takes longer than a pre-speci ed time interval. 10. When the hando is completed and the stability of the new link is established, the PZM deletes the user pro le and sends a RECORD UPDATE message to the user's permanent home database to update the user pro le. This message contains the user's IDN and the ATM address for the candidate zone.

Hando through the Candidate Port 1. The portable initiates a hando by establishing a link with the candidate radio port in the candidate zone and requesting a hando from the candidate zone

CHAPTER 3. SIGNALING IN THE WIRELESS ATM NETWORK USER

PZM

Request Handoff

CZM

36

END POINT HDB

HANDOFF_INIT

ACK. HANDOFF CHANNEL ASGN. HANDOFF REROUTE CHAN. INFO

REROUTE COMP.

REROUTE CONN.

VERIFY CONN HANDOFF COMP

HANDOFF COMP

RECORD UPDATE

Figure 3.10: Migratory Signaling Message Flow for Successful Hando, Case 1, Through the Previous Port manager. This message includes the user's identi cation number and previous zone identi cation. 2. The CZM will acknowledge this message. The CZM will send a HANDOFF INIT signaling message to the PZM. The HANDOFF INIT message contains the user's IDN and the address for the PZM and the CZM. The portable will tune back to its previous channel after the acknowledgment is received. 3. Upon receiving the HANDOFF INIT message the PZM will transfer the user pro le to the CZM using a HANDOFF INFO message. The PZM shall also send a HANDOFF REROUTE message to the end point for the user's connection. The rerouting is performed as speci ed in Section 3.2.2 and in Chapter 4.


37

4. If the rerouting is successful, the CZM shall provide the PZM with the channel assignment information for the portable using a HANDOFF CHANNEL ASSIGN message. It will also send a REROUTE COMPLETE message to the end point. If the rerouting is not successful, the hando attempt is dropped, the PZM and the end point are noti ed by sending a HANDOFF FAIL message listing rerouting failure as the reason. The PZM in turn noti es the portable. 5. If the re-routing is accomplished, the PZM sends the proceed with hando message to the portable that is still tuned to the channel in the previous zone. The proceed with hando message includes the channel assignment information for the candidate zone. If the portable cannot be reached at the previous zone, the CZM will be noti ed and the CZM will page the portable to establish a link. This page will also include the candidate channel assignment. 6. The portable, after receiving either the channel assignment information or the page from the CZM, will tune to the assigned channel and transmission is initiated. The CZM will send a HANDOFF COMPLETE message to the the PZM and the end point for the connection. The end point then will release the data that is being buered. At this stage the PZM releases the connection between itself and the end point[63]. 7. When the hando is completed and the stability of the new link is established, the PZM deletes the user pro le and sends a RECORD UPDATE message to update the user pro le to the user's permanent home database. This message contains the user's IDN and the ATM address for the candidate zone. The Simultaneous Hando discussed in Section 3.1.4, must also be considered in Migratory Signaling. If at least one of the parties involved in the hando perform an intra-zone hando, then the methods presented above will be sucient. If both parties involved in the hando perform an inter-zone hando then the previous zone managers (PZM) will be responsible for forwarding the signaling messages to

CHAPTER 3. SIGNALING IN THE WIRELESS ATM NETWORK USER

PZM CZM Init. Handoff

End Point

38 HDB

ACK HANDOFF INIT HANDOFF INFO HANDOFF REROUTE REROUTE CONN

Proceed

Verify Conn

REROUTE COMP and HANDOFF CHAN ASGN HANDOFF

COMP

RECORD UPDATE

Figure 3.11: Migratory Signaling Message Flow for Successful Hando, Case 2, Using Candidate Port

the candidate zone managers (CZM) for both users. The forwarding of signaling messages during a hando allows the wireless ATM network to be able to accommodate a simultaneous hando event by using the procedures discussed above. Since the signaling messages are routed correctly, the rerouting is not aected by the simultaneous hando.


39

3.2.3 Connection Setup using Migratory Signaling The connection setup procedure in the W-ATM network must be investigated in three stages:

Connections from wireless to wireless users. Connections from wireless to wireline users. Connections from wireline to wireless users. The connection setup from wireline users to wireline users is handled adequately by the current ATM signaling speci cations hence is not discussed here[6, 16]. Wireless users are mobile and change their locations. The user location information is updated by means of the registration process described in the previous sections. The main dierentiator between connections in the ATM network and the connections in the Wireless ATM network is the address or location resolution process performed before an actual connection is established; therefore, the standard ATM connection setup procedure becomes inadequate. We propose the following wireless ATM connection setup procedure for setting up connections in the wireless ATM Network. We consider the three cases mentioned previously:

Case 1: Connection Setup between Two Wireless Users We assume that the wireless users and the xed users are dierentiated by their user identi cation numbers. The ATM address for a xed ATM terminal is recognized as the IDN. For this case we will assume that wireless user A decides to call wireless user B. The procedure for connection setup is given below and illustrated in Figure 3.12: 1. User A's terminal contacts the zone manager for A's current zone. The terminal transfers the IDN for user B and the connection setup parameters to the zone manager. The zone manager for user A is referred to as the \ZMA" throughout the section.

CHAPTER 3. SIGNALING IN THE WIRELESS ATM NETWORK USER A

ZMA

Init. Conn.

Home DB for B ZMB

USER B

LOCATION REQUEST LOCATION REPLY MOBILE SETUP MOBILE ACCEPT

Channel Assgnmt Info.

40

Page User Ack.

MOBILE START USER CONNECTION

Figure 3.12: Migratory Signaling: Successful connection setup between two wireless users in dierent zones. 2. User B's IDN indicates that B is a wireless user. The ZMA checks the visitor location register to see if B is in the same zone. 3. If B is in the same zone as A then the ZMA allocates channels for both users and sends the assignment information to both A and B over radio links. This completes the connection setup process. 4. If B is not in the same zone as A then the ZMA starts a connection setup session. A connection setup session consists of two stages: The location resolution and the connection establishment. In the location resolution stage the ZMA nds the current location of user B. In the connection establishment stage the ZMA establishes the actual user connection between A and B. 5. Location Resolution: The ZMA uses B's IDN to resolve the ATM address of the permanent home database of B. The ZMA sends a LOCATION REQUEST message to B's permanent home database. The LOCATION REQUEST


41

message contains user B's IDN, the ATM addresses of the ZMA and B's permanent home database. The permanent home database of user B replies by sending a LOCATION REPLY message to the ZMA. The LOCATION REPLY is sent if and only if the calling user (A) is not in called user's (B) call blocking list. If user A is in the call blocking list, the permanent home database will send a REQUEST DENIED message to the ZMA stating the call blocking as the reason. The LOCATION REPLY message contains B's present location and the ATM address for the ZMA. The call blocking is checked at the home database for security reasons. 6. Connection Establishment: After retrieving the location information for user B, the ZMA contacts the zone manager of B's present zone (ZMB) by sending a MOBILE SETUP message. The MOBILE SETUP message contains the ATM address of ZMB, the user IDN for B and the connection parameters for the user connection. The intermediary nodes that receive this message reserve the appropriate resources and set up the virtual circuit translation tables for the connection. If an intermediary node is not able to allocate the required resources, it sends a MOBILE SETUP DENIED message to the ZMA, clearing the established circuits in the intermediary nodes on the path. If the MOBILE SETUP message reaches the ZMB, this means that the path throughout the network for the user connection is established and the resources are allocated. The ZMB pages user B. (a) If B responds to the page, then the ZMB allocates the radio channel for the user connection and sends this information to B. The ZMB then sends a MOBILE ACCEPT message to the ZMA and con rms the connection. (b) If B does not respond to the page then B is assumed to be inactive and the ZMB sends a MOBILE SETUP DENIED message to the ZMA, releasing the connection in the network. 7. If the user connection request is accepted by the ZMB, then the ZMA pages


42

A and relays channel assignment information. The ZMA sends a MOBILE START message to the ZMB indicating the start of user data. User A starts sending information.

8. If the user request is denied, the ZMA relays this information to A.

Case 2: Connection Setup from Wireless to Wireline Users For this case we assume that wireless user A decides to establish a connection to wireline ATM terminal B. B's ATM address is used as the user IDN for connection purposes. A connection from A to B is established as follows and is illustrated in Figure 3.13: User A

ZMA

Fixed Host B

Request Conn.

Ack. MOBILE SETUP

MOBILE ACCEPT Page User

MOBILE START

Figure 3.13: Migratory Signaling: Successful connection setup from a wireless to a wireline user.


43

1. User A establishes a radio link to the zone manager for A's current zone (ZMA). It requests a connection to be set up between A and B. The ZMA recognizes B's IDN as being a wireline ATM terminal. 2. Since the called user, B, is a wireline ATM terminal, there is no need for location resolution for this case. The ZMA constructs a MOBILE SETUP message to B. This mobile setup message contains: A's IDN, the ZMA's ATM address, B's ATM address and the connection setup parameters. 3. The MOBILE SETUP message sent by the ZMA is sent through ATM signaling virtual circuits. Each intermediary node that receives this message tries to allocate the resources required by the connection and sets up the translation tables. If the required resources could not be allocated for the connection, the intermediary node sends a MOBILE SETUP DENIED message to the ZMA clearing the connection at all the nodes on the path from itself to the ZMA. The MOBILE SETUP DENIED message cites insucient resources as the reason for denying the request. 4. If all the intermediary nodes between the ZMA and the wireline ATM host B are able to allocate the desired resources, the MOBILE SETUP message is received by B. B compares A's ID against its call blocking list. If A is in the call blocking list, B sends a MOBILE SETUP DENIED message. If A is not in the call blocking list, B sends a MOBILE ACCEPT message to A. This message is received by all of the intermediary nodes and virtual circuits are set up. 5. When the MOBILE ACCEPT message is received by the ZMA, wireless user A is paged. After A responds to the page, the ZMA sends a MOBILE START message to B. The connection is established. If A does not respond to the page, then the connection is cleared.


44

Case 3: Connection Setup from Wireline to Wireless Users In this case we assume that a wireline ATM terminal A requires a connection to be established to a wireless user B. The ATM address for B is equivalent to B's IDN. The connection setup proceeds as follows and is illustrated in Figure 3.14: Fixed Host A

Home Database for B ZMB

USER B

LOCATION REQUEST LOCATION REPLY MOBILE SETUP Page User Ack. MOBILE ACCEPT MOBILE START USER CONNECTION

Figure 3.14: Migratory Signaling: Successful connection setup from a wireline to wireless user. 1. The wireline ATM user A determines that B is a wireless terminal by the given ATM address for B. It then decides to employ the wireless ATM connection setup procedure to establish the connection. This proceeds in two stages. 2. Location Resolution: The wireline ATM user A resolves the ATM address of the permanent home database for B. A then sends a LOCATION REQUEST message to B's permanent home database. This message contains B's IDN, the ATM addresses of A and B's permanent home database. The permanent


45

home database of B replies by sending a LOCATION REPLY message to A. The LOCATION REPLY message is sent if and only if A is not in B's call blocking list. Otherwise, a REQUEST DENIED message is sent as a reply stating call blocking as a reason. The LOCATION REPLY message contains A's ATM address, B's IDN and the ATM address of B's current zone. 3. Connection Establishment: After receiving the LOCATION REPLY message from the permanent home database of user B, A sends a MOBILE SETUP message to the current zone of user B. This message contains A's ATM address, B's IDN, the ATM address of the current zone of B and connection parameters. We will refer to the zone manager of current zone of user B as the ZMB. All the intermediary nodes that receive the MOBILE SETUP message are required to allocate the necessary resources for the connection and setup the virtual circuit translation tables. If an intermediary node does not have the necessary resources, then it sends a MOBILE SETUP DENIED message to A clearing the connection. The MOBILE SETUP DENIED message cites insucient resources as the reason for denying the request. If the resources for the connection are available and the network path for the connection is established at the intermediary nodes, then the MOBILE SETUP message reaches the ZMB. Upon receiving the MOBILE SETUP message, the ZMB pages user B. (a) If B responds to the page, then the ZMB allocates the radio channel for the user connection and relays this information to B. The ZMB then sends a MOBILE ACCEPT message to the wireline ATM user A and con rms the connection. (b) If B does not respond to the page then B is assumed to be in inactive stage and the ZMB sends MOBILE SETUP DENIED message to the A thereby releasing the connection in the network. 4. If the user connection is accepted by the ZMB, A sends a MOBILE START message to the ZMB and the connection is established.


46

Clearing an Existing User Connection Clearing an existing user connection is performed in the same manner for all three cases listed above. The side that decides to terminate the connection sends a MOBILE CLEAR message to the other end point. The end point upon receiving the message sends a MOBILE CLEAR ACK message to the side that initiates the call clearing. All the intermediary nodes that receive the MOBILE CLEAR ACK message clear the connection. The end point sends the MOBILE CLEAR ACK message after all the outstanding data cells are received. By clearing the connection with the acknowledgment message the network ensures that there are no cells lost because of premature termination.

3.3 Estimated Signaling Overhead Required to Support Wireless Users Estimating the performance of a signaling network protocol requires the de nition of a performance measure. Our performance measure is the overhead required for signaling. We are not concerned with the signaling information being exchanged, but in the overhead required to carry that information. This performance measure is chosen because the signaling information needed to be communicated to perform network functions in any network remains roughly the same; however, the work performed in transmitting that information is a good measure of the eciency of that protocol. We used the methods given in [19] to estimate the signaling overhead required to support wireless users. These methods use the uniform uid ow model for approximating wireless user motion, and Poisson call arrival and exponential call holding time distributions given in [20]. The uniform uid ow model approximates the wireless users as moving randomly in all directions between the zones and on the average it is an acceptable representation of trac in and out of zones[19]. The Poisson call arrival and exponential call holding times are standard models for modeling telephone calls in conventional telecommunications networks. It is


47

also shown in Chapter 5 that Poisson call arrival models may be used to model connection arrivals in wireless ATM networks. Equations that yield mean values for the registration, Lr , connection setup, Lc, and hando, Lh, transaction rates in terms of transactions per second (TPS) are given in [19]. We have adapted and further simpli ed the equations given in [19] and arrived at the following set of equations (See Appendix B):

Lc = Npc(1 + q) ? m) N 4 Lr = (1 ? s)(1 T (1 ? s )(1 ? m ) 4 Lh = Ne T D T = v

(3.1) (3.2) (3.3) (3.4)

See Table 3.1 for the de nitions of the terms. The equations above yield results for any given con guration that are accurate estimates within the limitation of being mean values for statistical variables. . By examining the equations given above, the following are evident: 6

All of the transaction rates scale linearly with the number of people in a zone. Assuming that the population density remains constant, all of the above transaction rates vary proportional to the square of the zone size.

The rate of registration transactions is not aected by call arrival rate or call holding time.

The rate of connection attempts is not aected by the user velocities. Both the registration and hando transaction rates vary linearly with user velocities assuming that zone size and thus the number of people in the zone remain the same.

6

The validity of these equations is veri ed in [19].


48

Parameter Name Parameter Symbol Default Value Number of Users N None in a Zone Average Zone T None Crossing Time Call Arrival Rate c 3 calls/hour (Call Rate) Call Holding Time 180 seconds Velocity of users v 35 kmph Zone Size D 8000 meters (one side of the square) Penetration Ratio p 0.25 Successful Call q 0.5 Completion Stationary User s 0.50 Percentage Ratio of Mobile Users m 0.45 that stay in one zone Probability of a e c user being busy Table 3.1: Signaling Performance Estimation: Baseline Values used in the calculations.

We now proceed to discuss the performance estimates for both overlay and migratory signaling approaches. We note that the focus of the results are towards the eects of user velocity, calling rate and call holding time. Although we have performed all of the variations listed above, we only show a few examples of the linear relationships.

3.3.1 Performance of the Overlay Signaling Approach In this section we estimate the performance of the overlay signaling approach. For the estimates, we also specify the kind of ATM signaling needed to handle registration, hando and connection setup transactions. The number of ATM connections that are required to be established for overlay signaling, as noted in Section 3.1, are


49

Signaling Transaction Number of ATM Connections Registration (Explicit Deregistration) 2 Registration (Implicit Deregistration-Lazy) 1 Connection Setup (Mobile to Mobile) 3 Connection Setup (Mobile to Wireline) 1 Connection Setup (Wireline to Mobile) 3 Hando 4

Table 3.2: Overlay Wireless ATM Signaling Trac in terms of ATM connections that need to be established. Varied Parameter Values Call Arrival Rate 1-6 calls/hour ( Call Rate) Call Holding Time 60-360 seconds Velocity of users (Zone Size=8000 m.) 10-90 kmph Velocity of Users (Zone Size=16000 m.) 10-90 kmph Zone Size 4000-44000 meters Table 3.3: Signaling Performance Estimation: Varied Parameters. Zone size is 8000 meters unless otherwise stated.

summarized in Table 3.2. Based on these gures and the minimum and maximum number of bytes required to set up ATM connections (See Appendix A,[16]), we then calculate the mean number of transactions per second and the mean bandwidth required for setting up ATM connections to support these transactions. The parameters in Table 3.3 were varied one at a time to arrive at the results discussed in the text. Figures 3.15 and 3.16 show examples of the linear relationships. 1. The ATM Overlay signaling bandwidth required to support the wireless network is on the order of a few megabits per second for all the cases investigated. We conclude that the bandwidth required to support the zones of the wireless network is much less than the 155 Mbits/sec capacity of a typical ATM link and does not constitute a signi cant overhead.


50

2. The number of ATM connections that need to be established for supporting the wireless network mobility; however, is on the order of 100 to 400 connection attempts per second. We note that this gure does not include the user connections. This gure is a very signi cant overhead and translates to roughly one connection setup every 2:5 milliseconds. To the best of the author's knowledge this appears to be a challenge to current ATM switching technology. 3. All the gures given are for a square zone size of 8000 meters on one side. As the zone size increases, all of the quantities listed above either increase quadratically or linearly with the zone size. Thus, the overlay signaling paradigm does not scale well for supporting a large number of mobile wireless users. 4. Varying the call holding time by a factor three results in only a one percent change on the estimated ATM bandwidth required for signaling as shown in Figure 3.16. Hence, we conclude that the call holding time does not have signi cant eect on the ATM signaling bandwidth. Of course, it has a very marked eect on the bandwidth required to support user connections.

3.3.2 Performance of the Migratory Signaling Approach In this section we focus on the performance of the Migratory Signaling discussed in Section 3.2. We again use the mean number of signaling messages per second and the bandwidth required to transmit these messages as a performance criteria for Migratory Signaling. After performing the parameter variations listed in Table 3.3 we have observed the following: 1. Number of Signaling Messages sent and the bandwidth required to send them varies linearly with velocity, population of a zone and calling rate. Mean call holding time does not have a signi cant eect on the bandwidth or the number of signaling messages transmitted through the network.

CHAPTER 3. SIGNALING IN THE WIRELESS ATM NETWORK Velocity = 50 kmph

300

Attempts per Second.

51

Connection Registration Handoff ATM Connection

250 200 150 100 50 0

0

20

40 60 80 100 120 140 160 180 200 Number of People in thousands

Figure 3.15: Overlay Signaling Transaction Rates. Zone Size = 8000 meters, Velocity = 50 kmph 2. The mean number of signaling messages transmitted per second in the wireless ATM network varies between 20 and 400 messages per second depending on the population and other factors. 3. The mean signaling bandwidth required to support W-ATM signaling varies between 10 and 900 kilobits per second depending on the population size and other factors. These bandwidth gures were calculated using the proposed W-ATM signaling messages de ned in Appendix C. 4. The single most dominant factor for all measures of interest is the population of the zone. For a given population both mean calling rate and mean user velocity are equally important. The call holding time has no signi cant eect on signaling although its eect on user bandwidth is very important. Figures 3.17 and 3.18 show examples of the linear relationships. Based on these calculations we can now compare the Overlay Wireless ATM Signaling with Migratory

Bandwidth Kilobits per second


52

ATM Bandwidth Required for Overlay Signaling 1100 Call Holding Time = 60 sec. Call Holding Time = 180 sec. 1000 Call Holding Time = 300 sec. 900 800 700 600 500 400 300 200 100 0

0

20


Figure 3.16: Overlay Signaling Bandwidth vs. Call Holding Time Wireless ATM Signaling.

3.3.3 Comparison of the Overlay and Migratory Signaling Methods In this section we compare the Migratory Wireless ATM Signaling and the Overlay ATM signaling. The Overlay Signaling is designed to support wireless users in an ATM network with no modi cation to the existing ATM protocols and with only the addition of gateways to the ATM network. The Migratory Signaling is designed to be a native Wireless ATM protocol providing built-in support for both wireless and wireline users. To achieve built-in support, migratory signaling adds a few signaling messages to the existing ATM protocols. The comparisons follow:

The Overlay Signaling requires ATM connections to be established to support

mobile (wireless) users. In Section 3.1 we have shown that for varying population sizes, overlay signaling requires 40 to 250 ATM connections per second to

Signaling Bandwidth (Kilobits/sec)


53

700 Call Rate = 1 Call Rate = 3 Call Rate = 5

600 500 400 300 200 100 0

0

20


Figure 3.17: Variation of Required Bandwidth with Call Arrival Rate in Migratory Signaling be established resulting in the transmission of 200 to 1250 signaling messages per second. This presents a challenge to existing ATM switching equipment. The Migratory Signaling uses ATM signaling virtual circuits to transmit signaling message to support wireless users. Since these messages are sent using the existing ATM signaling virtual circuits, no connections need to be established. The number of mean signaling messages sent are on the order of 20 to 400 messages per second. This number is feasible using today's technology.

The bandwidth required to support signaling for wireless users is also an im-

portant criteria. Using our calculations we have found that the Overlay Signaling requires an average signaling bandwidth of 1:4 Megabits/second. For the same set of baseline parameters the Migratory Signaling requires an average bandwidth of approximately 500 Kilobits/sec.

W-ATM Signaling Messages per Second


54

600 Velocity = 10 kmph Velocity = 50 kmph Velocity = 90 kmph

500 400 300 200 100 0

0

20


Figure 3.18: Variation of Number of Messages with Velocity

The interaction between wireless and wireline users in the Wireless ATM net-

work is an important feature of our design. In the Overlay Signaling, this interaction is implemented by means of service gateways that function as intermediary points between the wireline and wireless users. In the Migratory Signaling, the support for wireless and wireline user interaction is built-in to the signaling protocol. This native support is also more ecient in terms of bandwidth and signaling routing since the connections need not be routed through the service gateways.

We have mentioned in Section 3.1 that the overlay approach may have a scaling

problem when the population of a zone exceeds 100; 000 people. The solution for this problem is further sub-division of zones with large population sizes. The Migratory Signaling is more robust in terms of scalability and it minimizes the scalability problem by using native signaling channels.


55

The Overlay Signaling requires no modi cation to the existing ATM protocols. The only ATM network modi cation required for Overlay Signaling is the establishment of service gateways in the networks as discussed previously. The Migratory Signaling requires addition of some signaling messages to the existing protocol to provide direct support for wireless users.

The ineciency associated with Overlay Signaling is a result of choosing to

use switched virtual circuits (SVCs) for signaling connections. An alternative to the use of SVCs is to use permanent virtual circuits (PVCs) for overlay signaling. By using PVCs, the zone managers of the wireless ATM network will be able to communicate without establishing connections. In such a network, the overhead associated with using PVC based overlay signaling will be comparable to migratory signaling discussed in this chapter. However, there are some disadvantages associated with PVC based signaling as noted previously and in [7]. The primary disadvantage is the need for manual administration of PVCs for each and every zone; moreover, the number of required PVCs will grow with the square of the number of zones in the network. Architectures given in [7] address the problems associated with PVC based overlay signaling. Also, most of these PVC based schemes require changes in the current ATM protocols. In contrast, SVC based overlay signaling does not require any changes in the existing protocols, it is completely distributed and requires no administration.

We also note here that these protocols may be used in tandem to evolve support for wireless users in an ATM network to form a \Wireless" ATM network.

3.4 Summary In this chapter, we describe two signaling protocols that may be used to integrate the wireline Asynchronous Transfer Mode and the next generation wireless personal communications networks into a wireless ATM network: Overlay Signaling is an


56

overlay protocol implementation to integrate mobility into the existing ATM protocol stack, and Migratory Signaling is a new signaling protocol that provides native support for both wireline and wireless ATM users by enhancing the existing ATM protocols. When we compare Migratory Signaling to Overlay Signaling, we observe that Migratory Signaling uses less bandwidth for signaling and does not need additional signaling circuits that are dedicated for support of wireless users. Overlay Signaling implements the support for wireline and wireless user interaction by using service gateways. Migratory Signaling provides built-in support for wireless users.

Chapter 4 Rerouting in the Wireless ATM Network Hando is implemented by the wireless ATM network to give the users freedom of motion beyond the coverage area of a single wireless access point. Hando is the procedure by which a user's radio link is transferred from one radio port to another through the network without an interruption of the user connection[14]. During hando, the user connection may be preserved using two alternative procedures: 1. Re-establishing the existing user connection: In this procedure, the user connection is re-established between the termination point of the connection in the wireless ATM network and the user via the new radio port. 2. Rerouting the existing user connection to the new radio port: In this procedure, the user connection is rerouted in the wireless ATM network. The connection does not need to be re-established. In this dissertation, we chose the second alternative due to the following reasons:

Re-establishing the user connection will result in a new connection setup and in related overhead as discussed in Chapter 3. 57

CHAPTER 4. REROUTING IN THE WIRELESS ATM NETWORK

58

Most handos in the wireless ATM network happen between neighboring radio

ports or neighboring zones due to the modes of transportation available to wireless users; i.e. walking or driving. Re-establishing a connection end-toend does not take advantage of the locality of most handos.

ATM protocols support multiple simultaneous connections per user[6]. Re-

establishing multiple user connections will signi cantly increase the connection setup overhead and delay.

In contrast to connection re-establishment, rerouting takes advantage of the locality of hando by extending or shortening the existing user connection during hando. The time required for rerouting the user connection will be less than the time required for re-establishing the user connection because of the locality of hando. Finally, multiple connections belonging to the same user may be grouped together and rerouted in the same step during hando. This chapter describes how ATM connections may be rerouted during hando in a wireless ATM network. We rst summarize the hando procedure. We then present a rerouting method based on cell forwarding for wireless ATM networks that use Overlay Signaling. Our motivation for using cell forwarding for Overlay Signaling is to remain compatible with existing ATM protocols while rerouting user connections. Rerouting for Overlay Signaling is described in Section 4.2.1. For wireless ATM networks that use Migratory Signaling, we develop the Nearest Common Node Rerouting (NCNR) protocol for rerouting of user connections in Section 4.2.2. NCNR is based on rerouting a user connection at the node that is the closest common ancestor of the zones that are involved in the hando. This allows NCNR to make maximum re-use of the existing user connection during rerouting and saves network resources as discussed later in this chapter. We conclude the chapter by comparing the proposed rerouting algorithms to the existing algorithms in the literature[17, 21, 22, 23, 24].


59

4.1 The Hando Procedure The hando procedure is performed to assure the integrity of a radio connection, to minimize interference to the users in the coverage area of neighboring cells and to gain access to better resource availability[14, 10]. The wireless ATM network consists of radio ports, user terminals and network interface equipment as described in Chapter 2. A wireless user might have multiple simultaneous connections in the wireless ATM network. When a hando occurs these connections may need to be rerouted. There are two levels in hando: Network Level and Radio Level. The radio level hando is the actual transfer of the radio link between two ports. The network level hando supports the radio level hando by performing rerouting and possibly buering. The radio level hando determines some of the procedures used in network level hando as we shall see later. We also assume that the zone managers have prior knowledge about the neighboring zones and the network addresses of neighboring zones are stored in a local lookup table which is updated periodically by means of an update protocol[25]. We refer to the user communication device as the user terminal and termination point for the user connection as the end point. In Chapter 3, we presented the methods by which a hando transaction may be implemented in a wireless ATM network using either the Overlay ATM signaling or the Migratory wireless ATM signaling protocol in the zone based wireless ATM network architecture. Based on the zone concept, a few dierent situations for the rerouting may be investigated:

A Hando within a Zone (Intra-zone Hando): In the intra-zone hand-

o the user is moving within the zone. The only rerouting that is performed in this case is in the equipment within the zone. Speci cally, the zone manager and the radio port controllers are responsible for the correct update of ATM virtual circuit translation tables within the zone; moreover, this type of rerouting does not require signaling and rerouting other than the functions performed by the zone. Due to the containment of hando within the zone, the intra-zone hando does not generate wireless ATM network signaling and is not discussed farther in this dissertation.


60

A Hando between Two Zones (Inter-zone Hando): The inter-zone hando occurs when the radio ports involved in the hando belong to dierent zones. In this case the rerouting involves the wireless ATM network and is coordinated by the two zone managers that are involved in the hando event. An inter-zone hando might require rerouting at one or more wireless ATM switches depending on the location of hando and the topology of the network. This type of rerouting will be discussed in the next section.

Wireless ATM users may subscribe to services ranging from time-sensitive traf c types (audio, video) to throughput dependent trac types (data, le transfers, world-wide web access). The trac type of the connection involved in the hando is known by the zone managers of the wireless ATM network. The two kinds of traf c types impose dierent constraints on the network and the hando process. For example, time-sensitive voice trac will not be easy to buer due to constant cell generation rate and strict time delay constraints; however, it can tolerate occasional loss of cells. On the other hand, data trac will not tolerate cell loss, but may tolerate delays on the order of few hundred milliseconds. The rerouting procedures for time-sensitive and throughput dependent trac must be dierent. We also note that, occasionally, a hando will be attempted without any warning due to severe fading in the radio environment. In such a case, the upper layer protocols will be responsible for recovery of user connection and lost cells. This case is analogous to cell loss in the xed ATM network due to severe congestion and should be treated in a similar manner[16].

4.2 Rerouting for Inter-zone Hando The rerouting for an inter-zone hando involves one or more wireless ATM switches and is handled dierently for Overlay and Migratory Signaling protocols discussed in Chapter 3. In this section we rst describe the rerouting for Overlay Signaling, then we propose a novel inter-zone hando procedure referred to as the Nearest Common Node Rerouting (NCNR) for Migratory Signaling.


61

4.2.1 Rerouting for Overlay Signaling Overlay Signaling for wireless ATM networks is designed to support wireless users in an ATM network without changing the existing ATM protocols as described in Chapter 3. The constraint of functioning within the boundaries of existing signaling protocols limits the rerouting functionality to two alternatives: Re-establishing the complete connection during a hando event or using the cell-forwarding techniques discussed later in this chapter. Since the wireline ATM network is using standard ATM signaling and the mobility functions are supported using Overlay Signaling, we can not use any of the rerouting techniques described in [67]. The rerouting that may be used must utilize only the zone managers and must use Overlay Signaling. A wireless ATM user may have multiple simultaneous connections to multiple end points in the wireless ATM network. Re-establishing multiple connections to multiple end points during an inter-zone hando will cause unnecessary signaling load and will not meet the quality of service requirements of the user connections due to slowness of re-establishing multiple connections simultaneously. The second option for rerouting using overlay signaling is cell forwarding. Cell forwarding is similar to the call forwarding approach currently being used in the telephone network. Cell forwarding also allows the wireless ATM network to optimize the rerouting procedure depending on the trac type.

Rerouting using Overlay Signaling for Time Sensitive Trac The cell forwarding procedure using overlay signaling for time sensitive trac in the wireless ATM network is described below and illustrated in Figure 4.1: 1. A hando is initiated between zones A and B managed respectively by Zone Manager for A (ZMA) and Zone Manager for B (ZMB). A is the present zone and B is the candidate zone as illustrated in Figure 4.1. 2. The wireless user's connection needs to be rerouted from zone A to zone B. Using the ATM Connection Setup procedure described in 3.1.1, ZMA establishes a bi-directional ATM connection to ZMB with trac characteristics


: Existing Connection ATM Switch

ATM Switch

: Forwarding Connection

ATM Switch

Candidate Zone (B)

Present Zone (A)

Previous Radio Port

Candidate Radio Port Handoff

Figure 4.1: Rerouting using Cell Forwarding

62


63

matching the wireless user's existing connection. This new connection is the cell forwarding connection. If the cell forwarding connection is not established due to resource availability then the hando event is terminated. 3. Once the cell forwarding connection is established, the rerouting in the wireless ATM network is completed and the radio level hando starts. 4. ZMA starts to forward the user information to ZMB using the cell forwarding connection. ZMA also passes the user information to the previous radio port in zone A that is in contact with the user(Figure 4.1). This results in a point to multi-point connection that is necessary until the radio level hando is stabilized. The radio level hando may extend beyond one radio burst in most radio systems as the user terminal tries to select the optimal link. Especially when a user is in a fading environment, a small motion of the terminal may cause the radio link to switch back and forth between the two radio ports involved in the hando; hence, a point to multi-point connection between the end point and zones A and B ensures the timely delivery of timesensitive information. If a zone has not received an uplink transmission from the portable in a given radio transmission frame, it must assume that it is not active for the next downlink radio transmission period (See [14, 10] and Chapter 3 for details on the hando procedure). The user information may then be discarded at the zone that is not in contact with the user terminal and transmitted from the zone that is in contact with the terminal. This zone is the zone that is currently receiving the uplink transmission from the portable. In the uplink direction, the information may be transmitted through either zone involved in the hando and correctly routed to the end point through ZMA. If a soft hando scheme such as in code division multiple access (CDMA) is being employed where the uplink information is being received simultaneously by two radio ports, then the ZMA is responsible for combining and sequencing the uplink information. Occasionally, the portable may receive duplicate information. The duplicate information may be determined by the time sequence information and discarded accordingly. The ATM transmission


64

and sequencing protocols for transmission of time-sensitive trac are currently being discussed in the ATM Forum ; hence, we will not go into speci cs of transmission of time-sensitive trac over ATM networks. 5. If the radio level hando is successful, ZMA acts as a regular ATM switch in the network and stops forwarding the user information to the previous radio port in zone A.

Rerouting using Overlay Signaling for Throughput Dependent Trac The rerouting of throughput dependent trac is very similar to the rerouting procedure employed for time sensitive trac with the following dierences: 1. As the radio level hando is started, the downlink user information is buered at both zones A and B. No user information is transmitted in the downlink direction until the radio level hando is completed. Once the radio level hando is completed, the information is transmitted in a rst in rst out (FIFO) manner. 2. If A's buer is non-empty before the hando is started, then A's buer is transmitted to the user terminal if possible; otherwise, these data are transmitted to B and go in front of all other cells buered for transmission. This preserves the cell sequence. 3. In the uplink direction, before the radio level hando is started, the trac is transmitted through A if possible; otherwise it is buered at the terminal. As the radio level hando is started, the user terminal starts buering the user information. Once the hando is stabilized the buered information is transmitted. These dierences ensure the integrity of user data as well as the cell sequence. The cell sequence aspects of rerouting are discussed in Section 4.3. If a simultaneous hando occurs as described in Section 3.1.4, the forwarding of signaling messages from the ZMA to the ZMB assures the success of rerouting.


65

4.2.2 Nearest Common Node Rerouting (NCNR) for Migratory Signaling The NCNR attempts to perform the rerouting for hando at the closest ATM network node that is common to both zones involved in the hando transaction. The term \common" is used to denote a network node that is hierarchically an ancestor of both of the zones in question (See Figure 4.2). Nearest common node rerouting Backbone ATM Network

Wireless ATM Network Node(s)

NCN for A&C

NCN for A&B Zone A

Zone B

Zone C

Figure 4.2: Depiction of Nearest Common Node minimizes the resources required for re-routing and conserves network bandwidth by eliminating unnecessary connections (See Section 4.4). We assume that due to the nature of the xed network, the transmission delay and latency of the links from the NCN to the zones involved in the hando are negligible compared to the radio transmission medium. In this section we rst explain the NCNR procedure for time(delay)-sensitive trac. We will then conclude by explaining the NCNR


66

procedure for throughput dependent trac based on the NCNR for time-sensitive trac.

NCNR for Time-Sensitive Trac (NCNR-TS) The NCNR for time-sensitive trac is performed as follows: 1. A hando session between zones A and B is started. Let B be the candidate zone for the hando. Let A be the present zone. 2. The zone manager of A rst checks to see if a direct physical link (not involving any other network nodes) between A and B exists. There are two possible cases if this condition is satis ed: Endpoint

WATM Node

B

A

WATM Node

WATM Node

Add-on User Connection

Candidate Port

Handoff

Previous Port

Figure 4.3: NCNR: Case 2.a in a Flat Network


67

Wireless ATM Network Previous Radio Port WATM Node A

WATM Node D

Portable Add-on User Conn. Handoff

WATM Node C

WATM Node B


Figure 4.4: NCNR: Case 2.a in Hierarchical Network (a) If A is a parent of B, then A noti es B and the new connection is established without any further network involvement. This case is illustrated in Figures 4.3 and 4.4. After the connection is established, A acts an anchor or a forwarding gateway for the connection. Until the stability of the hando is established, both A and B act as network connection points for the user connection. This process is explained in detail in step 6 of this procedure. Once the radio level hando is completed, then A acts only as a wireless ATM switch in the connection path. 1

2

The parent is determined by means of either the network topology in a hierarchical network or by means of closeness in terms of hops to the end point. The end point is de ned as the terminating point for the user connection in the network. 2 The stability of the hando is used in reference to the radio link transfer which may take longer than one burst during a hando event. During this period, the user terminal may use both network points for information transfer. 1


68

Endpoint

WATM Node

WATM Node A

WATM Node B

WATM Node

WATM Node

Deleted User Connection Previous Port Handoff

Candidate Port

Figure 4.5: NCNR: Case 2.b in a Flat Network (b) If B is a parent of A then A sends a message to B relaying the hando request. B then acts as an anchor for the hando procedure. Until the stability of the hando is established, both A and B may be used for information transfer from/to the terminal to/from the network (See step 6 of this procedure). Once the hando is stable, B deletes the user connection from itself to A. The rerouting is thus completed. This case is illustrated in Figures 4.5 and 4.6. 3. If A and B are not connected by a direct physical link then the zone manager of A (ZMA) contacts the end point for the user connection by sending a HANDOFF START message. The HANDOFF START message contains the ATM addresses of zones A, B and the endpoint for the user connection. 3

3

The signaling messages will be denoted by bold lettering in this chapter.


69

Wireless ATM Network Candidate Radio Port WATM Node B

WATM Node D

Handoff Deleted User Conn.

WATM Node C

WATM Node A

Portable

Previous Radio Port

Figure 4.6: NCNR: Case 2.b in a Hierarchical Network 4. The HANDOFF START message traverses the network from A to the endpoint for the user connection. The network switching nodes on this path upon receiving this message check to see whether all three ATM addresses are routed on dierent egress ports of the switch (See Figure 4.2). When such a node is found it is designated as the nearest common node (NCN). One exception to this test is when the endpoint for the user connection is also in zone B. This may be detected by examining the routing information for the endpoint of user connection and handled accordingly. The NCN sets the NCN bit in the HANDOFF START message. The rest of the switches on this path do not perform the egress port test. An alternative implementation would be not to forward this message after the NCN is found. Both are equally viable options. By forwarding the message to the end point we allow the user applications to adjust to the hando process.


70

5. The nearest common node then forwards a REROUTE message to all of the switches located between B and itself. The nodes that receive the REROUTE message rst check for resource availability; if the resources required by the connection are available, the necessary connections are established and circuit translation tables are set up. If the resources are not available, the hando attempt fails and the involved parties are noti ed. 6. When the REROUTE message is received by B, a REROUTE ACKNOWLEDGMENT message is sent from B to A. This message completes the rerouting process. 7. When the rerouting is completed, the radio level hando starts. NCN starts to forward the user information to both A and B in a point to multi-point manner. As described in Section 4.2.1, this multi-party connection is necessary until the radio level hando is stabilized. A point to multi-point link from the NCN to both A and B ensures the timely delivery of time-sensitive information. In the downlink direction, the information received by the zone that is inactive is discarded by the inactive zone and transmitted to the portable by the active zone. If a zone has not received an uplink transmission from the portable in a given radio transmission frame, it must assume that it is not active for the next downlink radio transmission period (See [14, 10] and Chapter 3 for details on the hando procedure). In the uplink direction, the information may be transmitted through either zone involved in the hando and correctly routed to the end point by the NCN. If a soft hando scheme such as in code division multiple access (CDMA) is being employed where the uplink information is being received simultaneously by two radio ports, then the NCN may be responsible for combining and sequencing the uplink information. If the portable receives duplicate information, the duplicate information may be determined by the time sequence information and discarded accordingly. 8. If the radio level hando is successful and the new radio link is stable, the connection between A and the NCN is cleared by A by sending a CLEAR


71

CONNECTION message to the NCN. After the hando is completed, any

buered time-sensitive data that has not expired will be transmitted to the current zone associated with the portable; expired data are discarded at the zones. For time-sensitive trac, recovery of lost data may only be possible by interpolation of information which is beyond the scope of this dissertation.

NCNR for Throughput Dependent Trac (NCNR-TD) NCNR for throughput dependent trac is very similar to the procedure employed for time-sensitive trac. Throughput dependent trac is not sensitive to small (on the order of few hundred milliseconds) delays; however, the loss of information is not tolerated by this trac type. A typical example of this trac type is a le transfer. We can take advantage of the delay tolerating nature of this trac in the rerouting process. The NCNR for throughput dependent trac diers from the procedure for NCNR-TS as follows: 1. As the radio level hando is started, the downlink user information is buered at both A and B. No user information is transmitted in the downlink direction until the radio level hando is completed. Once the radio level hando is completed, the information is transmitted in a rst in rst out (FIFO) manner. 2. If A's buer is non-empty before the hando is started, then A's buer is transmitted to the user terminal if possible; otherwise, these data are transmitted to B and go in front of all other cells buered for transmission. This preserves the cell sequence. 3. In the uplink direction, before the radio level hando is started, the trac is transmitted through A if possible; otherwise it is buered at the terminal. As the radio level hando is started, the user terminal starts buering the user information. Once the hando is stabilized the buered information is transmitted. These dierences maintain the integrity of user data as well as the cell sequence. The cell sequence aspects of rerouting are discussed in Section 4.3.


72

We also note here that multiple connections may be rerouted in the network using the virtual path concept. By assigning a virtual path identi er for connections between a user and multiple end points and performing the rerouting on the virtual path instead of on a virtual circuit basis, an ecient rerouting of multiple connections may be achieved. If a simultaneous hando as discussed in Section 3.1.4 occurs, the forwarding of signaling messages from ZMA to ZMB assures success of rerouting. The discovery procedure for locating the NCN is not aected by the simultaneous hando event.

4.3 Preserving the Cell Sequence in Rerouting In any ATM network, cell sequence of individual ATM cells must be preserved for correct re-assembly of encapsulated user information[16]. Therefore, during a hando, the preservation of cell sequence is a primary concern of the network. In a hando transaction, there are two parties involved: the hando terminal (HT) and the end point. In this section we will explain how the cell sequence is preserved in the rerouting algorithms discussed in the previous section.

4.3.1 Preserving the Cell Sequence using Overlay Signaling Overlay Signaling uses cell forwarding to manage the rerouting of connections in the wireless ATM network. Point to multi-point connections are used to provide robustness against the ping pong eect that may be observed during a a hando. Cell forwarding without point to multi-point connectivity preserves the cell sequence because all of the cells that reach the user follow the same path. Point to multi-point connectivity introduces a fork in the path where the cells are duplicated. Let us assume that the interarrival time between two cells in the connection to be rerouted is denoted by Tcell. jtC + tCP ? tPP j < Tcell (4.1) Using the notation given in Figure 4.7, if Equation 4.1 holds then the cells destined for the user arrive at the two radio ports approximately at the same time so that


t PP Previous Radio Port

73

Zone A tC Zone B

t CP


t C : Transmission delay between zones A & B t CP : Tranmission delay from zone A to previous radio port t PP : Transmission delay from zone B to candidate radio port

Figure 4.7: Timing for Cell Sequence Analysis the same cell is ready for transmission through both the candidate and the previous radio ports at the same transmission period. Since an assumption in our analysis is that the time latency dierence between the two zones involved in the hando is less than the period between the successive transmissions of time-sensitive trac, the above relationship always holds. For example, on a 10 Mbits/sec time sensitive trac stream , the 53 byte ATM cells, arrive approximately 4.24 msecs apart. A 155 Mbits/sec ATM link takes about 0.003 milliseconds to transmit 53 bytes. The wireline network latency variables tC ,tCP and tPP are all on the order of micro seconds whereas due to the lower available bandwidths in the wireless links Tcell will be on the order of milliseconds; hence, Equation 4.1 is always satis ed by a large margin. This fact, together with the cell discarding scheme mentioned in Section 4.2.1 helps preserve the cell sequence. For throughput dependent trac, no cells are transmitted to the user until the 4

10 Mbits/sec is a very high bit rate for a time-sensitive trac stream in a wireless ATM network. 4


74

radio level hando is completed (See Section 4.2.1). This removes the timing constraint described above and the only constraint for throughput dependent trac is achieving buer consistency for both zones. Since the present zone forwards all cells that may be buered at that zone at the beginning of hando to the candidate zone, the buers at both zones start in a consistent state and remain consistent due to the forwarding procedures described in Section 4.2.1. This demonstrates the preservation of cell sequence in rerouting using Overlay Signaling for both time-sensitive and throughput dependent trac types.

4.3.2 Preserving the Cell Sequence using NCNR and Migratory Signaling Nearest Common Node Rerouting (NCNR) is used in Migratory Signaling to perform rerouting in the wireless ATM network. For time-sensitive information the cell sequence in NCNR is preserved due to the fact that the cells in the connection stream are duplicated at the NCN and transmitted to both zones involved in the hando. The transmission delay in the network compared to the time separation of the radio bursts is negligible ; hence, the cells arrive at approximately the same time at both zones. This could be proven using analysis similar to Equation 4.1 in Section 4.3.1. The zone that is currently active with the HT transmits the cell and the other zone discards it. This reduces the number of active paths to one in the downlink direction, thereby preserving the cell sequence. In the uplink direction, data from only one zone are forwarded to the endpoint by the NCN. This preserves the cell sequence. If a soft hando scheme is being used where the uplink direction data are received by two zones simultaneously, then NCN will be responsible for combining the two streams into one. For throughput dependent trac, the zone managers start buering information from the start of radio level hando to the completion (successful or unsuccessful). Also note that until the radio level hando is started, the user connection is assumed 5

E.g. The radio bursts in PACS are separated 2.5 milliseconds apart[10], where as transmission time of a 200 bit burst takes on the order of 1.5 microseconds on a 155 Mbits/sec ATM link. 5


75

to be active and all the incoming cells are transmitted to the HT through the previous port. If that is not possible, the cells are forwarded to the candidate port through the NCN for later transmission. In the reverse direction, the HT buers the user information as soon as the radio hando is initiated. The cells that are in transmission in the network from the HT to the end point are delivered normally. By employing the procedure discussed above for throughput dependent trac, the cell sequence during the rerouting of a user connection due to a hando attempt is preserved. At a given time during a hando, there is only one active path between the two parties involved in the connection. Since there is only one active path at any given time, all the cells transmitted through the network take this path and arrive in sequence to their destination. Occasionally, cell sequence will be broken in the wireless ATM Network. Recovery from such sequencing errors is the responsibility of upper layer protocols such as the ATM Adaptation Layer (AAL) and sometimes the user application layer. For data trac, recovery may be attempted by retransmission; however, for time-sensitive trac the sequence error may need to be corrected by voice or video interpolation techniques.

4.4 A Comparison of Existing Rerouting Algorithms for Wireless ATM Networks The cell forwarding based rerouting for Overlay Signaling and the Nearest Common Node Rerouting for Migratory Signaling enhance the current ATM networks by adding support for rerouting an active connection. The support for rerouting is necessary for supporting mobile users. In the current version of ATM Signaling Protocols there is no provision for rerouting a connection once it is established[16]. In this section we compare the rerouting approaches proposed in this dissertation with other rerouting algorithms proposed for wireless ATM networks. These will be referred to as Yuan-Biswas([21]), BAHAMA([22]), Virtual Connection Tree (VCT) ([17]), SRMC([23]) and CSDR([24]). We will rst explain these rerouting


76

algorithms then compare them to the rerouting algorithms proposed in this dissertation. Finally, a quantitative comparison of these rerouting algorithms will be presented.

4.4.1 Description of Alternate Rerouting Approaches Yuan-Biswas Rerouting Algorithm In this wireless ATM rerouting algorithm, the wireless ATM connections are rerouted at designated hando switching equipment (HOS). The hando procedure described in [21] does not specify how the HOS is determined. The example in that text uses the initial wireless ATM switch as the HOS. This in turn is very similar to cell forwarding. It is assumed that the base stations in the wireless ATM network are interconnected by permanent virtual circuits. A hando between two ports (or base stations) attached to the same wireless ATM switch is handled by just updating translations tables in one switch. A hando between two ports attached to dierent ATM switches is handled by forwarding the ATM cells destined for the user terminal to the user's new ATM switch. This new add-on connection is established before hando is completed. The cell sequence is preserved since the rst switch acts as a hando server (switch).

BAHAMA Rerouting Algorithm for a Wireless ATM LAN The BAHAMA hando algorithm is proposed for a wireless ATM local area network (LAN). The BAHAMA architecture consists of a at network of radio ports (base stations) and user terminals. The radio ports are interconnected with ATM links. The BAHAMA architecture is also based on forwarding of cells after a successful hando. The initial ATM switch acts as an anchor (forwarding switch) for the hando connection and forwards the user cells to the new radio port. After the hando is completed the initial ATM switch migrates the user connection to an optimal route in the network provided that the portable stays within the coverage area of the new radio port for an extended period of time. Since the cells are always routed through the initial switch during hando, the cell sequence is preserved. The


77

BAHAMA LAN uses the virtual path indicator of the ATM cell header for routing. This simpli es the rerouting since all that needs to be changed in the cell header is the virtual path indicator provided that the virtual connection indicator is available at the new radio port.

VCT: Virtual Connection Tree Based Rerouting Algorithm In the VCT hando algorithm the concept of a virtual connection tree is utilized. A virtual connection tree is formed by a root node that is attached to the backbone ATM network. The nodes that are connected to that root node form the tree structure. When a mobile terminal establishes a wireless ATM connection, a connection tree is formed from the root node to the leaves in the tree eectively producing a point to multi-point connection; however, the mobile terminal is utilizing only one leaf node at a time whereas the rest of multi-point connection is not utilized. When the mobile moves within the tree, a new leaf node becomes active and the connection is continued using one of the pre-established circuits. When a mobile moves out of the coverage area of the virtual connection tree, the network establishes a new tree surrounding the mobile.

SRMC Rerouting The Source Routing Mobile Circuit (SRMC) rerouting approach is an improvement of the VCT rerouting. In this approach the rerouting functions are distributed over time. The SRMC approach uses the concept of a tethered point (TP) to serve as the root in the connection tree for hando. When a connection is rst being established, all potential network routes from the TP to the leaves, due to possible hando attempts are identi ed by the network and these connections are pre-established. Unlike the VCT algorithm, no resources are reserved. Once the hando is initiated, the resources for only the active hando connection are reserved. After the completion of the hando, the TP is possibly migrated and new possible network routes are determined. The dierences between SRMC and NCNR will be discussed in Section 4.4.2. The reader is referred to [23] for details on this algorithm.


78

Crossover Switch Discovery Rerouting (CSDR) CSDR is based on the concept of rerouting by nding an optimal crossover node in the network to perform the rerouting[24]. The crossover node is similar to the Nearest Common Node discussed earlier in the text. The optimality is de ned in terms of meeting quality of service requirements. This approach requires the control of a centralized node for crossover node discovery[24]. Once the crossover node is discovered, the rerouting is performed in a manner similar to NCNR proposed in this dissertation; however, it does not distinguish between time-sensitive or throughput dependent trac types. CSDR is proposed for the local area network (LAN) environment. For handos between LANs, CSDR uses boundary nodes that act as gateways. CSDR is capable of handling multicast connections.

4.4.2 Comparison of Rerouting Algorithms In this section, we compare and point out the dierences between the rerouting approaches discussed in the previous section. One key aspect that needs to be observed is whether the proposed algorithms require changes to the existing ATM signaling protocols. Rerouting for Overlay Signaling proposed in this dissertation is the only rerouting algorithm that does not require changes to the existing ATM protocols. It handles time-sensitive and throughput dependent trac types. Hence, Rerouting for Overlay Signaling is a unique rerouting protocol and will not be compared with the rerouting algorithms discussed previously. Discussions presented below that relate to cell forwarding represent to some extent Rerouting for Overlay Signaling. For wireless ATM networks, where a modi cation of the signaling protocols is possible, we propose Nearest Common Node Rerouting. In the following sections, we compare NCNR to alternate rerouting algorithms proposed in the literature.

Comparison of NCNR with Yuan-Biswas and BAHAMA hando algorithms In this section we compare NCNR with the Yuan-Biswas and BAHAMA rerouting algorithms that may be employed for hando in the wireless ATM network. The


79

rst point to note is the cell forwarding aspect . The main reason for the use of cell forwarding in the network is the ease of cell re-sequencing. When the ATM switch associated with the previous port acts as an anchor for the rerouting, all the cells still go through the previously established path in the network before traversing the new add-on section established because of a hando. This guarantees the preservation of cell sequence and also means that either the anchor ATM switch([22]) (the hando switch([21])) or the new ATM switch might need to buer some of the cells before the hando is successfully completed. For only one user this might not be a signi cant administrative load; however, the simulations in Chapter 3 predict hando rates of 10 or more handos per second in a cellular environment. In such a scenario the buering of cells due to hando may become a burden to the network. In NCNR the buering is mostly performed for throughput dependent trac and only when the radio level hando is being performed. Moreover, for time-sensitive trac, buering is not feasible, and the Yuan-Biswas and BAHAMA rerouting algorithms do not allow for supporting a user connection through two radio ports while the hando stabilizes. This may ultimately cause a problem to time-sensitive trac streams. Another problem with the cell forwarding model is that it inherently assumes a

at (or a ring) network model where all neighboring ATM switches (or zones) will have direct connections in between. The reason for this assumption is that in a at network the cell forwarding does indeed minimize the number of ATM switches that are involved in a hando. However, in a hierarchically organized network, the cell forwarding based rerouting automatically involves the node that is referred to as the \nearest common node (NCN)" in this dissertation(See Figure 4.2). Once the NCN is involved in cell forwarding it is more advantageous to use the NCN Rerouting (NCN) algorithm proposed herein. This is primarily because of two points: 1. When the NCN is involved in the hando, the network bandwidth is actually minimized when the connection is simply rerouted to the new ATM switch (zone) of the wireless network. This prevents the waste of bandwidth for the 6

The YUAN algorithm has provisions for dynamic rerouting of connections that are similar to NCNR; however, presently the details for the dynamic rerouting are not available. 6


80

portion of the forwarding connection between the NCN and the previous ATM switch (zone) . 7

2. Utilization of NCNR means that the data for all the connections involved in the hando are buered at the zones (See Section 4.2.2). Buering of data is hence performed at the edge of the network requiring no intervention from ATM switching equipment . NCNR also preserves cell sequence as explained in Section 4.3. 8

Another possible problem with cell forwarding is the fact that for a fast moving user in a system that has small radio coverage areas, there will be a network trail left behind the user consisting of cell forwarding among multiple zones. NCNR always performs the rerouting at the nearest common node thereby minimizing the amount of bandwidth used and minimizing the amount of rerouting. The BAHAMA algorithm uses virtual path indicators of the ATM cell header for routing purposes in the wireless LAN. This approach, while suitable for a local area network, will not scale well for a wider area network. NCNR is developed with wide area networks in mind, and does not have such a problem.

Comparison of NCNR with the Virtual Connection Tree Based Hando Algorithm In this section we compare NCNR to the Virtual Connection Tree based hando proposed in [17]. The virtual connection tree based hando is similar to the Nearest Common Node Rerouting. There are three primary dierences: 1. NCNR takes the concept proposed by VCT one step further by deleting the need for multiple virtual circuits to be reserved at a given time to support a single connection. The zone concept utilized in this dissertation is eectively equivalent to the connection tree concept given in [17]. The zone manager The BAHAMA algorithm allows for reorganizing a connection based on cell forwarding once the user becomes stationary within the coverage area of a radio port. 8 Note here that the radio controller equipment may co-exist with the switching equipment. 7


81

node is the root and the radio ports attached to the zone manager node are the leaves. Since the radio ports and the zone manager maintain constant communication there is no need for multiple virtual circuits. 2. NCNR guarantees the preservation of cell sequence. The virtual connection tree based hando does not implement cell sequence preservation. The mobile is responsible for cell sequencing. 3. NCNR recognizes the diering constraints associated with time-sensitive and throughput-dependent trac types and implements the hando procedure to accommodate both types of trac in an ecient manner. The virtual connection tree based routing described in [17] does not address this issue. Because of these reasons, NCNR may be superior to the virtual connection tree based hando.

Comparison of NCNR with the SRMC Algorithm The SRMC algorithm improves the VCT hando algorithm by addressing most of the concerns that were expressed in the previous sections. SRMC does not reserve bandwidth in the connection tree until the actual rerouting is performed. This clearly is an advantage over the VCT algorithm. As compared to NCNR; however, we can still point out these unaddressed issues:

SRMC, by pre-determining all possible hando paths from a root node, at-

tempts to avoid the actual connection establishment during hando. It does not, however, avoid the resource allocation that needs to be performed before the actual rerouting is completed. This resource allocation process involves sending a message from the root node (TP) in the connection tree to all the nodes that are on the active path. There are also messages that are sent from the leaf node to the root node to notify the TP of an ongoing hando attempt. In summary, there is a pair of messages that are sent for noti cation and for resource allocation from one side to the other side of the connection tree. Since the root node has to be involved in all hando attempts, even an


82

attempt between neighboring nodes is managed through the root node. When we compare this with NCNR we see that the worst case in NCNR degenerates to SRMC in terms of number of messages that are sent between the nearest common node and nodes involved in the hando. When the nodes involved in the hando are neighbors, then NCNR outperforms SRMC since the messages go up only one level in the hierarchy. Therefore, in terms of number of messages sent between the nodes in the network hierarchy NCNR is at worst comparable, and, for neighbors, better than SRMC.

SRMC uses the centralized intelligent network (IN) concepts for pre-determining the possible routes that may be involved in a hando. When a user does not perform a hando, then all of the overhead of calculating these possible routes is wasted. It is also not clear whether the most resource intensive part of hando rerouting is resource allocation or nding a possible route for rerouting. The NCNR on the other hand, is a fully distributed algorithm. It performs the work only when it is necessary, avoiding wasted computational overhead.

SRMC does not address the constraints associated with dierent trac types. (See previous discussion in this chapter.)

SRMC inherently assumes a hierarchical topology. The algorithm is not eective in a at network.

It remains to be determined whether NCNR or SRMC will have the best performance in terms of speed of hando. However, in terms of eciency we believe that NCNR will indeed outperform SRMC for the reasons stated in this section.

Comparison of NCNR with CSDR The CSDR algorithm has been proposed in [24] for a wireless local area network (LAN) environment. This algorithm is similar to the algorithm discussed in this dissertation, but is implemented for wireless LANs. CSDR was implemented using the Cambridge Fairisle ATM Switch in a LAN environment and the results in terms of hando rerouting rates that may be accommodated in a LAN environment are


83

certainly encouraging. The main dierences between CSDR and NCNR are listed below:

NCNR is proposed for a wide area wireless network and requires minimal mod-

i cations to current ATM signaling speci cations using the migratory signaling proposed in Chapter 3.

NCNR does not require any buering to be performed in the network switching

nodes. The buering of data during a hando is performed at the zones for throughput-dependent information only, avoiding a buer management problem as mentioned in the previous sections. Buering can not be used for time-sensitive information as noted earlier.

NCNR provides support for dierent types of trac, and provides point-tomultipoint support for handling time-sensitive trac streams in a hando event. No distinction for dierent trac types is made in the CSMR.

NCNR is a fully distributed rerouting algorithm that involves only a partial assembly of the nodes in the connection path. CSDR is a centralized rerouting algorithm that relies on the knowledge of the network topology. While this may be suitable for a wireless LAN environment, centralized control of a wide area network will have scalability problems.

NCNR algorithm will function equally well without any intermediary or boundary nodes in the network. CSDR relies on boundary nodes for rerouting of wireless ATM connections between LANs. The management of boundary nodes may actually undermine the optimization algorithms used in CSDR.

Based on the discussion presented above, CSDR is a good rerouting protocol for the LAN environment, but it is not suitable for rerouting of wireless ATM connections across wide area networks.


84

4.4.3 Quantitative Comparison of Rerouting Algorithms Quantifying rerouting algorithms presents a challenge. There are two possible approaches. One approach to quantifying a rerouting algorithm is extensive simulation of a wide area network either on a hardware level or by using network simulators. A second approach is to characterize the algorithm and simplify the measurement approach. In this dissertation we chose the second approach for characterizing and quantifying various rerouting algorithms. In general the following may be of interest for comparing hando and rerouting algorithms: 1. Signaling Bandwidth used in Hando: The signaling bandwidth used in a hando event may limit how much bandwidth is available for the user connections; therefore, for the procedures that achieve the same functionality , the amount of bandwidth used for signaling is a valid performance measure. Unfortunately, this aspect of the other hando protocols is not discussed in the literature and hence can not form a part of our comparison. 9

2. Number of signaling messages exchanged during hando (Nh): The number of messages involved in the hando procedure increases the complexity of the signaling software and may slow the hando process; therefore, the smaller the number of messages, the easier the hando procedure is to implement, and the faster it should run. 3. Number of signaling messages exchanged for rerouting during a hando (Nr ): The number of signaling messages due to the rerouting may slow the hando process and increase the complexity of the signaling software; therefore, the smaller the number of rerouting messages, the easier the hando procedure is to implement and the faster it should run. In order to be able to compare dierent hando procedures, we need to de ne a common functional description that is to be expected from a hando and rerouting protocol. If, for example, we were to design a circuit, we may compare two designs that meet the functional description of the circuit. The same concept holds true for network protocols. A common functional description is needed as a reference point. 9


85

4. Number of network nodes involved in the rerouting (Nn): This criterion determines the time it takes to perform the rerouting and the bandwidth required for signaling. 5. Number of user connections established for rerouting (Nc): This parameter is signi cant for virtual connection tree based hando procedures and determines the amount of overhead the network has to perform in order to achieve hando. 6. User Bandwidth allocated for hando (B ): This is expressed as a multiple of the basic bandwidth allocated for the user connection. In some virtual connection tree based hando procedures, the amount of user bandwidth allocated for a user connection may be a multiple of the actual needed bandwidth. 7. The time it takes to execute a hando: This measure is greatly in uenced by the underlying radio technology of the network. Hence two procedures that run on dierent radio networks may not be compared directly. (The CSDR algorithm proposed by Toh has been implemented by using the Cambridge Fairisle switch and hando times on the order of tens of milliseconds have been reported[24].) 8. Whether the protocol is proposed for wide (W) or local area networks (L). 9. Rerouting Algorithm used: Cell Forwarding (CF), Dynamic as in NCNR (DY), Tree based as in VCT or SRMC (T). 10. Robustness to Instabilities in Hando: The instability in a hando event refers to the ping-pong eect that sometimes occurs when the user terminal is in an area where the radio link to both radio ports may fade. In such a situation, the terminal may switch back and forth between the two radio ports a few times until one of the links stabilize. Hence, a hando procedure should be robust with respect to this eect. We will subjectively rate the respective hando algorithms on a scale of 1 to 5 with 5 being the most robust in our comparison based on our perception of the algorithms. The most robust algorithms take the ping-pong eect into account and are designed accordingly to minimize


86

additional network trac due to the ping-pong eect, the least robust algorithms are prone to generating additional network trac due to the ping-pong eect. The algorithms rated 3 are in the middle of the scale in both aspects. We can use the criteria above to compare the various rerouting approaches discussed in this dissertation with the exception of Rerouting for Overlay Signaling. Rerouting for Overlay Signaling uses the overlay signaling approach for rerouting connections and should not be directly compared to rerouting algorithms that rely on an enhanced version of current ATM protocols. The results of our analysis are given in Table 4.1. Symbols in Table 4.1 are de ned as follows:

x: Not described in literature. N: Number of leaves in the tree structure in virtual tree based algorithms. D: Number dependent on network topology. Measure

Nh Nr Nn Nc B

W/L Rerouting Algorithm Robustness Estimated by the author

NCNR

Direct No Direct Link Link 7 9 2 4 2 4+D 1 1 1 1 W W DY DY 5

5

Yuan-Biswas BAHAMA

VCT

SRMC CSDR

12 x 2 (est.) x 10 4 2 (est.) 2 N Centralized 6 2+D 2+D (est.) N + D N +D 3+D 1 1 N N 1 1 1 N 1 1 W L W W L CF / D Y CF T T DY 2

3

4

2

Table 4.1: Comparing Rerouting Algorithms This table may be used as a rst step in choosing a hando and rerouting algorithm for a given wireless network.

3


87

4.5 Summary In this chapter we described how rerouting may be implemented in a wireless ATM network. We have introduced a rerouting method based on the cell forwarding concept for wireless ATM networks using Overlay Signaling. We also described Nearest Common Node Rerouting (NCNR) for wireless ATM networks using Migratory Signaling. NCNR is based on nding the ATM node that is a root of or common to both of the zones involved in the hando transaction. The rerouting is then performed starting from this common node. We presented a survey of other rerouting algorithms from the literature and compared our rerouting procedures with the alternatives. This survey and comparison showed that NCNR is a good algorithm for rerouting wireless ATM connections. Rerouting for Overlay Signaling is the only rerouting protocol that works with existing ATM protocols as de ned in [6] and is advantageous in this regard. Another contribution of our research on rerouting for wireless ATM networks is the recognition of dierent constraints associated with the hando of dierent trac types.

Chapter 5 Medium Access in the Wireless ATM Network The wireless ATM network architecture proposed in this dissertation provides wireless access to a backbone ATM network. In order to be compatible with the backbone ATM network, the radio ports in the wireless ATM network need to employ a medium access protocol that supports multiple trac types with dierent priorities and quality of service requirements. Conventional medium access control protocols such as ALOHA, CSMA and SCRMA are designed to support packet data trac in wireless data communication networks and can not accommodate multiple trac types due to their design[26, 27, 28]. In recent years, dierent medium access protocols for wireless ATM networks have been proposed in the literature[18, 29, 30, 31]. However, these protocols have been either limited to speci c environments such as indoor wireless local area networks([29, 31]) or have been tied to a speci c radio technology[18, 30]. Dynamic Resource Allocating Multiple Access (DRAMA), described in this chapter, is a novel medium access control and resource allocation protocol that supports multiple users, multiple connections per user and service priorities. DRAMA is compatible with existing ATM protocols and is designed to operate in multiple environments. DRAMA is not tied to a speci c radio technology or speci c set of radio frequencies. The results of our simulations indicate that DRAMA performs better 88

CHAPTER 5. MEDIUM ACCESS IN THE WIRELESS ATM NETWORK

89

than or equal to previously proposed medium access control protocols for wireless ATM networks while providing the added bene ts discussed later in this chapter. In this chapter, we rst describe the wireless ATM radio and network environment as it pertains to medium access control and resource allocation. We then discuss the DRAMA algorithm and conclude by presenting simulation results that evaluate the performance of DRAMA and compare it with other resource allocation algorithms that have been proposed in the literature.

5.1 Wireless ATM Network Environment DRAMA is a dynamic resource allocation and medium access control algorithm for a multi-tier wireless ATM network. A multi-tier wireless ATM network is de ned as a network that may provide its users with varying data rates depending on geographic location. For example, data rates in an indoor environment may be close to Ethernet speeds at 10-20 Mbits/sec[11, 32]. In a campus environment, it may be feasible to provide a data rate of 2-8 Mbits/sec as described in various wireless ATM network proposals[33, 34, 35]. In a wide area network, the user may still gain access to the bene ts of the wireless ATM network but at a lower rate of 256-512 Kbits/sec[10, 36]. These three environments are referred to as Local Area, Campus and Wide Area environments respectively. The need for a multi-tier network arises because of the scarcity of radio spectrum and power available for radio transmission in a portable unit. The higher data rates for the campus and local area environments are likely to be provided in the 2 Ghz and higher frequency bands. At those frequencies, the range of radio transmission is limited by walls, partitions, etc[10, 14]. Due to this fact and to the scarcity of radio spectrum in the 800-900 Mhz frequencies, a multitier network is needed[34, 36]. DRAMA is designed with multi-tier wireless ATM networking in mind and is able to accommodate multiple types of ATM trac and to provide quality of service guarantees in resource allocation. It also bene ts from statistical multiplexing and allows data users full access to the channel bandwidth on a burst basis when full channel resources are available. DRAMA is designed to function in a time-slotted and frequency multi-plexed


90

radio environment as shown in Figure 5.1 (Figures for Chapter 5 are located at the end of the chapter). In such an environment, the users can access multiple time slots in a single frequency or multiple time-slots spanning multiple frequencies. The time-slotted and frequency multi-plexed radio channel is represented by an M by N channel matrix where M is the number of slots in each frequency and N is the number of total frequencies assigned to each radio port (See Figure 5.2). In the DRAMA environment, the downlink frequencies are only accessed by the radio ports and the access is strictly controlled by the radio port controllers. There is no contention on the downlink radio channels; hence, we limit our discussion in this dissertation to the uplink frequencies and aim to resolve the contention that is caused by the user requests on the uplink channels. A resource allocation that uses multiple time-slots spanning multiple frequencies is de ned as a frequency-time sliced allocation. The optimal resource allocation policy for accommodating dierent trac types in a time slotted-frequency duplexed environment has been shown to be frequency-time sliced allocation[37]. Therefore, in DRAMA we chose to use a frequency-time sliced resource allocation (RA) algorithm. A practical limitation associated with frequency-time sliced resource allocation is the frequency switching time required by most of today's radio transceivers. A radio transceiver will require a non-zero time to switch between two frequencies during transmission or reception of a data burst if a frequency-time sliced resource allocation technique is used. This creates blind slots in the channel matrix (See Figure 5.3)[38]. This problem is represented by a Frequency Switching Constraint on the resource allocation algorithm. It is possible to formulate an optimal allocation problem that allocates requests in a frequency-time sliced manner for multiple requests by performing an optimization that includes both existing and queued requests and also considers the frequency switching time limitations described above. Due to the nature of the radio trac which has relatively short connection durations and relatively high mobility, it is not feasible in most cases to solve this optimization problem on the order of tens of milliseconds. In most networks, the radio port controller needs to perform this allocation in a period of milliseconds or faster while possibly involving hundreds of users and hundreds of time slots in tens


91

of frequencies. The DRAMA algorithm performs successful resource allocation in an environment with frequency switching constraints using a heuristic greedy resource allocation algorithm that is described later in this text. As mentioned above, DRAMA is designed to support multiple trac types with dierent service quality and bandwidth requirements. The existence of dierent trac types with dierent service criteria requires the resource allocation algorithm to set priorities. For the purpose of performance estimation, CBR trac has priority over data trac. Note that DRAMA is capable of working with dierent priority policies including time of expiry, or quality of service based priority schemes. For requests within the same priority class, we employ a rst-come- rst-served (FCFS) priority scheme for allocating resources to requests. Please refer to Chapter 3 for details of the wireless ATM network architecture.

5.2 Dynamic Resource Allocating Multiple Access Algorithm Gaining access to a network resource can be divided into two tasks: Requesting the resource from the resource broker and allocation of the resource by the broker. As discussed in Chapter 2, the wireless users are connected to the wireless ATM network via radio ports. The radio ports are controlled by radio port controllers. Each radio port controller may control more than one radio port. The radio ports are designed to be small and economical radio modems that are easy to deploy. The intelligence of the radio network is in the implementation of the radio port controller and the switching hardware. In this section we present the DRAMA algorithm in two parts: Medium Access Control (MAC) and Resource Allocation (RA).

5.2.1 Medium Access Control in DRAMA Medium Access Control is de ned as the procedure of gaining access to a particular shared medium for the purpose of information transmission. For example, the IEEE 802.3 standard is a medium access control protocol for local area networks based on a


92

shared medium bus topology (ethernet). The radio channel is the shared medium in the wireless ATM Network. Gaining access to the shared radio channel is a two step process when the DRAMA algorithm is employed. The rst step is to send an access request to the radio port controller. The access request is sent in the designated signaling and control channel(s) (See Figure 5.1). These signaling channels may be shared by all three tiers of the wireless ATM network. Our analysis of the DRAMA algorithm is based on a time-slotted transmission format for the signaling channels. The signaling channels are accessed using a slotted-ALOHA algorithm. The slot format of the signaling channel is dierent from the user channels which are considerably wider in bandwidth. Existence of a single signaling channel for access to multiple frequencies and multiple tiers of the network allows the network to bene t from statistical multiplexing on the signaling channel in contrast to having a smaller number of signaling slots in each user frequency. With the advent of \soft" radio technologies the added cost of receiving a considerably narrower signaling channel together with the wideband user information channels is minimal[39, 40]. When a user terminal needs to send information through the wireless network, it accesses the next available slot in the signaling channel and sends a short request detailing the type of service requested (See Figure 5.4). A collision occurs when more than one user terminal access the same signaling slot to send a request . If there are no collisions in the slot; the request is received by the radio port controller and processed by the RA algorithm which will be discussed soon. If there is a collision, the user terminal does not get an answer for its request from the radio port controller and the request times out. For a timed out request, the user terminal employs an exponential backo procedure for retrying the request[41]. If the user terminal is not successful in the maximum allotted time for medium access it may try a dierent radio port or may report failure to the upper layer protocols. In a multi-tier environment it is bene cial to use the same control signaling formats for all of the tiers in the network for seamless connectivity while crossing network tier boundaries; hence, all three tiers in the wireless ATM network use the 1

We will not consider the capture eect which may only improve the results discussed in this paper. 1


93

same control messages. Due to the dierent trac parameters in dierent network tiers, signaling channel bandwidths may be dierent among the tiers.

5.2.2 Resource Allocation in DRAMA Before discussing the details of the resource allocation part of the DRAMA algorithm, we digress here brie y to de ne the Channel Chunk Matrix (CCM). The Channel Chunk Matrix is used while performing resource allocation. The CCM is related to the channel matrix de ned in Section 5.1. The CCM is an M by N matrix created by the port controller for indexing and referencing the available wireless ATM network resources. The channel chunk matrix is formed by grouping and counting the empty slots in each frequency and sorting them in the order of magnitude. For example, for a frequency channel of 6 slots, if the slots are (Idle,Idle,Idle,Busy,Idle,Idle), the associated column of the chunk matrix will be ((3,0),(2,4)), where (3,0) denotes that there is an available chunk of 3 slots starting at position 0 in that frequency; similarly, (2,4) denotes a chunk of 2 empty slots starting at the 4th position (See Figure 5.2). By using the channel chunk matrix, the radio port controller can allocate the available resources rapidly. We discuss the resource allocation procedure next. The resource allocation process in DRAMA is as follows: 2

1. Sort and Combine the requests: The signaling channel consists of time slots that are grouped into signaling frames(See Section 5.2.1). The radio port controller scans the received slots at the end of each signaling frame. All successfully decoded requests in a frame are combined with requests that were saved from previous frames. The combined requests are sorted with respect to priority class and order of arrival. The sorted requests are ready for further processing. Requests that were queued from previous frames get higher priority in the same priority class. 2. The resource allocation is then performed on a request by request basis (See 2

Each frequency is represented by a column of the channel matrix.


94

Figure 5.5): (a) Check Resource Availability: The request size is compared with the total number of available slots in the channel. If the request size is greater than the total number of available slots and the trac type is CBR , the request is queued; otherwise, the algorithm advances to the next step. (b) Exact Match: In this step, the port controller scans the channel chunk matrix for a match of the same size as the request. If an exact match is found, then the controller checks to see if frequency switching constraints (See Section 5.1) are satis ed for that user. If the constraints are satis ed, the resources are allocated, the channel and channel chunk matrices are updated; otherwise, the process is repeated until an exact match is found or all frequencies in the channel matrix are completely scanned for an exact match. If this step is unsuccessful, the port controller advances to the next allocation step. This procedure is illustrated in Figure 5.6. (c) Bigger Chunk: In this step, the port controller scans the channel chunk matrix for any chunk that is larger than the request size. When such a chunk is found, it is checked for frequency switching constraints. If the constraints are satis ed, the chunk is allocated; otherwise this process is repeated until a feasible chunk is found or all frequencies in the channel matrix are completely scanned. If a feasible chunk is not found after all possible frequencies are scanned then the port controller advances to the next allocation step. This procedure is illustrated in Figure 5.6. (d) Same Frequency: In this step, the port controller aims to allocate the user request in a single frequency to avoid the blind slot problem caused by frequency switching. It uses the channel chunk matrix to determine whether a frequency with enough slots exists. If such a frequency exists then the available slots in that frequency are checked for frequency 3

4

Data requests can be partially allocated; therefore we need to check the trac type The time that is required to switch between two frequencies determines the frequency switching constraints. Since the DRAMA algorithm supports multiple connections per user, these constraints need to be checked at every stage of the resource allocation process. 3 4


95

switching constraints. If the constraints are met then the slots are allocated, otherwise the port controller scans the next frequency with enough available slots. This step is repeated until a feasible frequency is found or all available frequencies are scanned. This procedure is illustrated in Figure 5.7. (e) Multiple Frequencies: Finally, the port controller attempts to perform a multiple frequency assignment. First, the port controller determines the frequency with the highest number of available slots. Then all feasible slots in that frequency are allocated; the remaining size of the request is determined by subtracting the number of slots already allocated from the request size. The multiple frequency allocation is repeated until the remaining size is zero (the request is completely allocated) or all frequencies in the channel matrix are completely scanned. The allocation step is unsuccessful when all frequencies in the channel matrix are completely scanned but the request can not be allocated. The multiple frequency allocation procedure is illustrated in Figure 5.8.

The procedure discussed above will nd a feasible allocation with the least amount of frequency switching. Since there are only M slots in a frequency, no portable can ask for more than M slots in a time frame.

3. The procedure given in the previous step is repeated for all of the valid requests. At the end of the allocation process, all portables with pending requests are noti ed of the success or failure. We note here that failed requests may be queued for later allocation until the request expires. The expiration time of a request is de ned by the trac type and the user application(s).


96

5.3 Performance Estimation of the DRAMA Algorithm The DRAMA algorithm is designed to perform resource allocation and medium access control in a wireless ATM network. A discrete-event simulator that is written in C programming language running on SUN Sparc and Intel Pentium based workstations using Solaris 2.5 and Linux was used to test the eectiveness of the DRAMA algorithm. In the following sections we describe the simulation environment and discuss the performance measurements that were performed. A presentation of the parameter variations and the results of these parameter variations follow the performance measurements.

5.3.1 Simulation Environment The simulations discussed in this chapter present an implementation of the DRAMA algorithm in a single radio port, multiple user environment. We chose to implement a single radio port in order to isolate the eects of interference from neighboring ports from the eects of the DRAMA algorithm. The co-channel interference from neighboring radio ports due to frequency reuse may be mitigated by interference cancellation and by dynamic channel allocation methods[42, 43]. Mitigation of cochannel interference during resource allocation must be addressed in future research. We assume that there are N user frequencies and a signaling frequency allocated to the radio port. We assume that both user and signaling frequencies are time slotted frequency multi-plexed links. On the downlink, the radio port is the only transmitter; hence we limit our analysis of the DRAMA algorithm only to the uplink environment. Each of the N user frequencies consist of M time slots (See Figure 5.2). In order to accurately implement the medium access control portion of our simulator, we speci ed the time slot format for multiple environments, the signaling frequencies for the environments and the time required to switch from one frequency to another in terms of time slots. During the beginning of a simulation run we can, at will, isolate Medium Access


97

Field Name Number of Bits User ID 64 Trac Type 8 Number of Slots 10 Request ID 32 Optional 46 Cyclic Redundancy Check 32 Total 192 Table 5.1: The Format of the Request Slots Control from Resource Allocation by turning the collision detection o. This allows us to isolate the eects of medium access control from resource allocation. A unique feature of DRAMA is the fragmentation of data across time. For example, if a user has a large packet to transmit, the user sends in only one resource allocation request. If the request can not be accommodated in one user data frame, it is divided in time and transmitted in subsequent frame(s). This increases the throughput of the system and avoids unnecessary retransmissions on the signaling channel. The fragmentation of data may be turned on or o during the start of the simulation. The format of the user requests is given in Table 5.1. The duration of a request slot is determined by dividing the total size of a request by the request channel bit rate. The frame period of the request channel is an input parameter.

Trac Models for Wireless ATM Networks Since no wireless ATM networks exist, modeling the user trac types and arrival patterns for wireless ATM networks is a challenging problem. Excellent reviews of trac modeling for wireline data communication networks are found in [44, 45, 46]. Trac models for voice telephone networks have been published and used for many years[20, 19, 47]. For traditional voice telephone networks, the Poisson Process is widely-accepted as a valid arrival model for voice calls[19, 47]. For wireless voice networks, the


98

Poisson process is still a good representation of the trac arrivals. The Continuous Bit Rate trac in the wireless ATM network represents applications such as voice communication, video conferencing , and possibly distance learning. All of these applications are initiated by people and have relatively long durations; hence, as in xed voice networks, we assumed that the CBR requests from the users arrive according to a Poisson process with exponentially distributed connection durations. There are three types of CBR trac diering in arrival rates, requested bandwidth and connection durations. We determined the bandwidth demands of CBR trac using bit rates required for voice, video conferencing, and distance learning applications. We also used previously published studies for determining baseline parameters for our trac models[29, 30, 31]. Modeling the data trac in a wireless ATM network is challenging. Self similar trac models exist for modeling data trac on an aggregated physical link such as an ethernet network[48, 44, 45, 46]. Brie y, self similarity is de ned by Mandelbrot as having the same autocorrelation function independent of the time scale being used[49]. For example, fractals that are often observed in nature show self similarity in space. For detailed discussions on self similarity the reader is referred to [48] and references therein. Self similar trac models use a time-slotted model and it is empirically con rmed that the number of user data packets in a given time slot is self similar. In [44], it is observed that the self similar nature of the data trac arises from the high variability of individual data sources. In particular, if a data source has ON and OFF periods, and either the transition time from ON to OFF or the transition time from OFF to ON are chosen from a distribution with nite mean and in nite variance (heavy tailed) then the aggregate trac observed over the link shows the self similar behavior discussed above. The Pareto distribution, illustrated in Equation 5.1, is an example of a heavy tailed distribution when is less than 2. To illustrate the dierence between a heavy tailed and an exponentially distributed random variable, we plotted two time series with the same mean but one drawn from the Pareto distribution and one from the exponential distribution in 5

5

Such as xed rate compressed video.


99

Figures 5.9 and 5.10. Note the compactness in the dynamic range of the exponential distribution. (5.1) F (x) = 1 ? ( x +x ); x > 0 For our wireless ATM network simulator, we utilized two dierent models for data trac. Our rst model employed a heavy tailed Pareto distribution as given in Equation 5.1 for determining the inter-arrival time of individual packet trains. Our second model employed a Poisson arrival model for determining the packet train inter-arrival times. The length of the packet trains was generated using a discrete uniform distribution. The heavy tailed interarrival times were used in environments where the bit rate available to the user was suciently high so that we can use models generated for Ethernet networks. In environments where the available bit rate was low, we used a Poisson arrival model for individual packet trains since we can not justify using a model that was generated for Ethernet networks. In order to cover both extreme possibilities in Campus Wide and Local Area Network environments, we also used Poisson arrivals as an alternate arrival model for data trac. The results for Poisson arrivals are compared to the Self Similar trac model results in the related sections. We have tested the performance of the DRAMA algorithm using three dierent tiers of the wireless ATM network and a wide range of parameter variations in all tiers. The dierent tier environments are discussed next.

5.3.2 Simulation Parameters Three dierent tiers of the wireless ATM network were considered in the DRAMA simulation:

Wide Area Network: In this environment the users were limited to low bit rate CBR and low bit rate data trac. Arrivals for both trac types were Poisson distributed with dierent rates.

Campus Wide Network: In the campus environment, the users may utilize all three CBR and also high bit rate data trac types. Since the available bit

CHAPTER 5. MEDIUM ACCESS IN THE WIRELESS ATM NETWORK 100 rate was relatively high in the campus wide network, we simulated the data trac arrivals using both the heavy tailed Pareto distribution given in Equation 5.1 and exponentially distributed inter-arrival times. The CBR arrivals were assumed to be Poisson. Use of a heavy tailed distribution in determining inter-arrival times results in a self similar aggregate trac pattern for data requests. Exponentially distributed packet inter-arrival times result in a Poisson arrival process.

Local Area Network: In the local area we concentrated mainly on data trac

and some high bit rate CBR trac such as video conferencing. The data trac was simulated using both self similar trac and Poisson arrival models similar to the campus wide environment.

See Table 5.2 for details on these environments. The following measures are used to evaluate the performance of the DRAMA algorithm:

Measures Related to Resource Allocation:

{ Failure Rates: Requests that were successfully received by the radio port controller but no resources were allocated within the time-out period. The failure rates are determined by dividing the number of failed requests by the total number of requests of that trac type. { Average Access Delay: The time interval between the time a request is made by a user terminal and the response to the request is sent by the radio port controller. The average access delay includes the time spent in resource allocation. { Fragmentation Ratio: Fragmentation Ratio is the percentage of data requests that got fragmented over two or more user data frames for transmission. This parameter is more important in CWN and LAN environments where there is more data trac.

Measure Related to Medium Access Control:

CHAPTER 5. MEDIUM ACCESS IN THE WIRELESS ATM NETWORK 101

{ Collisions: Percentage of requests that collide while accessing the signaling channel using the slotted-ALOHA algorithm.

These measures capture the most important aspects of the performance of the DRAMA algorithm and allow us to compare DRAMA with alternative medium access control schemes proposed in the literature. Variable LAN Campus WAN Number of Users 50 50 100 Number of Frequencies 3 5 10 Number of Slots 64 64 8 End Time 36000 10800 86400 Signaling Channel Slot Width 5120 bits 512 bits 256 bits Signaling Channel Frame Period (slots) 4 16 8 Arrival Rate 0.1 0.009 0.0014 Trac Ratio 0.25 0.50 0.75 CBR LOW Bit Rate - 32 Kbits/s 32 Kbits/s CBR MED Bit Rate 320 Kbits/s 128 Kbits/s CBR HIGH Slots 1.4 Mbits/s 256 Kbits/s Maximum Packet Size 40960 bytes 2048 bytes 256 bytes Frequency Switching Time 1 slot 1 slot 1 slot Data Timeout 1.0 sec 1.0 sec 1.0 sec CBR Timeout 5.0 sec 5.0 sec 5.0 sec Duration of Low Trac 180.0 180.0 180.0 Duration of Medium Trac 360.0 600.0 180.0 Duration of High Trac 600.00 600.0 180.0 Ratio of Low Trac 0 0.6 1.0 Ratio of Medium Trac 0.9 0.3 0 Ratio of High Trac 0.1 0.1 0 Table 5.2: Baseline Values for Simulation Environments

Based on the three tiers mentioned above, the following parameter variations about the baseline values were performed:

Number of Users served by the radio port


Arrival Rate Number of user frequencies available to the radio port Number of Slots in one frame in a user frequency Number of Slots in the signaling frequency Variable LAN Campus WAN Number of Users 30 - 160 10 - 80 10 - 500 Number of Frequencies 1-6 1 - 10 1 - 30 Number of Slots 32-96 24 - 96 8 - 48 Number of Signaling Slots 2-8 8 - 32 4 - 32 Aggregate Arrival Rate (1/sec) 0.07 - 0.4 0.005 - 0.009 0.001 - 0.005 Per User Table 5.3: Summary of Performed Parameter Variations

In order to quantify the eects of the variations (See Table 5.3) that were performed we needed a uni ed measure of the demand on the system. This parameter is usually referred to as the Oered Trac. The oered trac is an average quantity that represents the demand on the system resources. By looking at the oered traf c and the performance results we can compare the performance of various resource allocation algorithms or various parameter variations in a uni ed manner. In order to estimate the average oered trac we need to know:

The arrival rates of dierent trac types. The average number of slots per frame occupied by these trac types. The average duration that is spent transmitting user information. The average oered trac per user is calculated by multiplying the average arrival rate (i) by the average duration to determine the oered trac in units of Erlangs. The oered trac in Erlangs is converted to slots per frame by multiplying the previous result by the average number of occupied slots per frame for that trac

CHAPTER 5. MEDIUM ACCESS IN THE WIRELESS ATM NETWORK 103 type (si). Summing over all trac types as indicated in Equation 5.2 yields the average oered trac for the system in terms of slots per user data frame. The average normalized oered trac for the system, denoted by TOffered?Normalized is calculated by multiplying the average oered trac per user by the number of users and dividing the result by the total available system resources where M is the number of slots per frequency and N is the number of frequencies in that radio port. N TOffered?Normalized = (i i Mi sNi) Nusers (5.2) =1

Equation 5.2 is obtained by applying Little's result from queuing theory[47].

5.3.3 Simulation Results In order to evaluate the performance of the DRAMA algorithm we have performed hundreds of simulation runs. The results of the simulations were plotted in over a hundred gures. In this section, we summarize our results and show representative gures. We should also note here that since each environment uses a dierent set of baseline values and trac mixes the results are not comparable between environments.

Wide Area Network The wide area network represents a low data rate environment similar to current cellular phone environments, but having an added packet data capability. The trac mix is roughly 75 percent CBR and 25 percent data trac (See Table 5.2). The results that we obtained are summarized below:

Results Related to Medium Access Control:

{ There are no request failures due to the medium access part of the

DRAMA algorithm. All request failures are due to the non-availability of channel resources in resource allocation. The collision rate on the signaling channels is much less than 1 percent for all ranges of oered trac. Since the percentage of collided requests is less than 1 percent for all

CHAPTER 5. MEDIUM ACCESS IN THE WIRELESS ATM NETWORK 104 ranges of oered trac, the access delay directly related to the collisions in the signaling channels is negligible in overall access delay calculations.

Results Related to Resource Allocation:

{ CBR Failure Rates are less than 1 percent up to 83 percent normalized

oered trac. This corresponds to approximately 375 users supported by 10 channels of 256 Kbits/sec bit rate. Since the wide area network environment is quite similar to the conventional telephone networks, it is possible to compare the performance of DRAMA with call blocking observed in a conventional telephone network from statistics predicted by the Erlang-B trac formula[47]. Using the same baseline parameters and assuming that the blocked calls are immediately cleared, a conventional telephone network with 80 channels experiences roughly 2.5 percent blocking with 375 users using formulae given in [47]. We observe that the CBR failures observed in DRAMA are better than the call blocking observed in conventional telephone networks. This result is not surprising once we note that in DRAMA blocked calls are retried until a pre-determined timeout period expires whereas in Erlang-B blocked calls are immediately cleared. This explains the dierence between the two results. If the timeout periods in DRAMA are set to zero, then the CBR failures in DRAMA are very similar to the call blocking statistics observed in conventional telephone networks. If we use the M/M/m/K/N nite population, m-server and nite storage queuing system given in [47] where K denotes the maximum queue length and N denotes the number of users, we obtain a CBR failure rate of 0.1 percent for the same baseline parameters and 375 users. This result is much closer to the results predicted by DRAMA since the nite population nite queue length system is a better model for DRAMA than the blocked-calls immediately cleared Erlang-B model. { Data Failure Rates are less than 4.1 percent up to 83 percent oered trac at the signaling channel rate of 16 Kbits/sec.


{ The average CBR access delay is less than 40 msecs up to an oered 6

trac of 80 percent.

{ The average data access delay is less than 50 msecs up to an oered

trac of 70 percent. Since CBR trac has higher priority than data , data trac experiences more delay. As described above, the data access delay is also directly related to the resource allocation process.

{ The data fragmentation ratio is plotted in Figure 5.13 for dierent signaling rates. Less than 20 percent of all data requests are fragmented in time up to an oered trac of 80 percent. As expected, the fragmentation of data requests is not aected by the signaling rate since fragmentation is only aected by available resources.

See Figures 5.11, 5.12 and 5.13 for details.

Campus Wide Network The campus wide network is a relatively high bit rate environment that supports all three kinds of CBR trac and high bit rate data trac. The trac mix is 50 percent data and 50 percent CBR trac (See Table 5.2). The results of our simulations are summarized below:


{ The collisions on the signaling channel are less than 20 percent at all

ranges of oered trac for a signaling channel bit rate of 32 Kbits/sec. No requests fail due to collisions on the signaling channels; that is, all retries are successful. Refer to Figure 5.14 for details. The access delays due to the collisions do not constitute an important part of the access delays since most requests are retried only once.

Results Related to Resource Allocation: 6

The value of the average access delay is dominated by the signaling frame length


{ The observed CBR Failure rates are less than 1 percent up to 88 percent

oered trac as seen in Figure 5.15. In that gure we can see that up to 90 percent oered trac, the failure rates observed using both 16 and 32 Kbits/sec signaling are very similar. At oered trac conditions higher than 95 percent, 32 Kbits/second signaling seems to result in slightly fewer CBR failures. However, if we examine the 95 percent con dence intervals plotted for CBR failures in the same gure, we observe that the con dence interval for the data includes the data points for both 16 and 32 Kbits/sec signaling. Con dence interval calculation is discussed later in this section. Based on this gure, the variations observed in Figure 5.15 for oered trac higher than 95 percent are not numerically signi cant enough to justify a conclusion that 32 Kbits/sec signaling is better than 16Kbits/sec signaling in this environment.

{ The data failure rates are less than 1 percent up to 90 percent oered trac as seen in Figure 5.16.

{ The CBR and Data access delays were less than 10 msecs up to an oered

trac of 80 percent with a signaling channel bit rate of 32 Kbits/sec. With 16 Kbits/sec signaling the delays are approximately twice as long for that range of oered trac. This can be observed in Figures 5.17 and 5.18 and is a direct result of increased signaling frame length due to the lower signaling bit rate. At higher ranges of oered trac, the access delays grow rapidly due to queuing in resource allocation for both 16 and 32 Kbits/sec signaling. The observed delays also get closer.

{ Last Fragments delays are plotted in Figure 5.19. The last fragment de-

lays are directly aected by the signaling frame length which determines how often the queued requests are processed. The signaling frame length is directly related to the signaling channel bit rate. Due to this fact, higher signaling rates result in lower last fragment delays as can be observed in Figure 5.19. When the oered trac is low, only a few requests

CHAPTER 5. MEDIUM ACCESS IN THE WIRELESS ATM NETWORK 107 are fragmented, this results in high variability in the observed last fragment delays. The observed results fall within the 95 percent con dence intervals as calculated later in this section.

{ The data fragmentation ratio for both 16 and 32 Kbits/second signaling channel bit rates are plotted in Figure 5.20. The slight dierence between the data fragmentation ratios between 60 and 100 percent for the two signaling rates can be attributed to the dierence in the collision rates as shown in Figure 5.14. As the oered trac exceeds 100 percent, the system becomes overloaded and the two data fragmentation curves get closer.

See Figures 5.14 through 5.20 for details. We have also simulated the same Campus Wide environment using Poisson arrivals, an example of failure rates with Poisson arrivals is plotted in Figure 5.21. Looking at this gure, we immediately observe that DRAMA protocol handles bursty self similar trac better than the more regular Poisson trac streams. We also observe that there are fewer collisions on the signaling channel with Poisson arrivals, also the internal queue sizes, when data trac with Poisson arrivals is used, are about half the size of the queues required for handling self similar trac. In summary, the radio port controllers in the wireless ATM network do more work when accommodating self similar trac but the nal results are better. The queue buers in the radio port controllers for self similar trac need to be increased to accommodate the bursty arrival patterns. The results shown for the CWN environment are signi cantly better than the results shown for the WAN environment in the previous section. The dierence can be explained by noting that in CWN environment there are more data requests and data requests multiplex better than CBR requests with long durations. Secondly, the CWN environment has more available resources and this results in increased trunking eciency for the resource allocation algorithm. The increased trunking eciency also results in fewer queued requests and in reduced access delays.


Local Area Network In an indoor environment there may be other network alternatives for CBR trac; therefore, we chose the local area environment to be dominated primarily by data trac (See Table 5.2). The trac mix in the local area network is 25 percent CBR and 75 percent data trac. The results we obtained for the local area network are summarized below:


{ The collisions on the signaling channels are more pronounced in this environment with an observed collision rate of 7 percent at 100 percent oered trac with a signaling rate of 64 Kbits/sec as shown in Figure 5.22. The collision rate is inversely proportional to the signaling rate as observed in the gure. All retries of collided requests were successful; that is, no requests fail due to collisions on the signaling channel. Since the percentage of collided requests is small and since most collided requests succeed in the second try, the eect of collisions on the access delays discussed below are negligible.

Results Related to Resource Allocation:

{ There are no CBR failures up to 90 percent oered trac, with failures

much less than 1 percent of total attempted CBR requests up to 100 percent oered trac as illustrated in Figure 5.23. { The data failures are less than 1 percent up to 100 percent oered trac as illustrated in Figure 5.24. The variations observed in Figure 5.24 for high oered trac are within the con dence intervals as discussed later in this section. { The CBR access delay in the local area environment is less than 20 msecs in all ranges of oered trac as shown in Figure 5.25. { The data access delays are less than 45 msecs up to an oered trac of 85 percent as shown in Figure 5.26.


{ Fragmentation of Data: As described in section 5.3.1, the data re-

quests may be divided across time to improve the throughput of the system. As illustrated in Figure 5.27 , at 100 percent oered trac the fragmentation of data was at 35 percent (35 percent of all data requests got transmitted across multiple user data frames). We also observed that in almost all instances the request was divided only into two parts. The delay that was observed for the last fragment of a data request is shown in Figure 5.28.

As mentioned in the previous sections, we have also simulated data trac using Poisson arrivals in the Local Area Network environment. Poisson data arrivals cause more failures for data requests but also result in lower delays. Self similar trac requires longer internal queues in the radio port controllers of the wireless ATM network. Overall, the results support a conclusion that DRAMA handles burstier self similar trac better than less bursty Poisson trac patterns but more processing and buering in the radio port controllers of the wireless ATM network are required for burstier trac patterns. See Figures 5.22 through 5.27 for details.

Con dence Intervals We have calculated 95 percent con dence intervals for the simulations discussed in this chapter. A 95 percent con dence interval suggests that approximately 95 percent of all independent simulations that are run using the same input parameters should produce results that fall in that con dence interval. We digress here and discuss how a con dence interval is calculated. Let us assume that we have observed \n" independent samples of a random variable denoted by X . Let xi denote the value of sample number i, let x denote the sample mean and let S denote the standard deviation of the n samples. A (1 ? ) percent con dence interval for the expected mean (E [X ]) of this random variable is determined by 2

n x = nxi 1

(5.3)

CHAPTER 5. MEDIUM ACCESS IN THE WIRELESS ATM NETWORK 110 n S = ((nxi??1)x) s xlower = x ? z Sn 2

1

2

2

s

xupper = x + z Sn 1 ? ' P (?z N (0; 1) z): 2

(5.4) (5.5) (5.6) (5.7)

The expected mean is then predicted to be between xlower and xupper for (1 ? ) percent of all random samples of the same variable. \z" is calculated from Equation 5.7 where N (0; 1) denotes a Normal random variable with zero mean and a variance of 1. In our case in order to obtain a 95 percent con dence interval, we set to be equal to 0:1. For discussions of calculation of con dence intervals in dierent simulation environments please refer to [50]. In order to illustrate the con dence intervals found in our simulations, we have plotted some example con dence intervals in Figures 5.29 through 5.32. 2

5.4 A Comparison of Medium Access Control and Resource Allocation Algorithms for Wireless ATM Networks In this section, we present alternative medium access control and resource allocation algorithms for wireless ATM networks that have been proposed and analyzed by other researchers. Comparisons of these algorithms with DRAMA are also presented. The leading alternatives to DRAMA are Multiple-Services Dynamic Reservation Protocol (MDR) from NEC([30]), Distributed Queuing Request Update Multiple Access (DQRUMA) from Lucent([29]) and Polling Multiple Access (PMA) from Carleton University of Ontario, Canada([31]). All of these algorithms use either reservation or polling for access to shared radio resources unlike ethernet where terminals listen for collisions before they transmit. This is due to the following reasons:

CHAPTER 5. MEDIUM ACCESS IN THE WIRELESS ATM NETWORK 111 1. Collision detection based schemes do not work eectively in radio environments where some of the terminals may not hear the transmissions of other terminals. This is referred to as the infamous \hidden terminal" problem[26, 27, 28]. 2. In a radio environment where radio resources are typically highly loaded (offered trac close to 1), collision detection based schemes suer from fairly large access delays as they wait for the channel to become available. Reservation or polling based medium access schemes avoid both of these problems at the expense of added network access delay. In the following sections, we will brie y summarize the alternate multiple access schemes for wireless ATM networks and present a comparison of the alternatives with DRAMA.

5.4.1 Multiple-Services Dynamic Reservation (MDR) versus DRAMA Multiple-Services Dynamic Reservation Protocol was the rst algorithm to be proposed for medium access control in wireless ATM networks. It uses a 2 Mbits/sec, single frequency wireless radio link that is divided into request mini-slots and user data slots. MDR supports multiple trac types and can oer quality of service guarantees. The medium access control in MDR is similar to DRAMA. The users access the request mini-slots and transmit a request in the uplink direction. Once the resources are allocated, the users transmit their data. The resource allocation portion of the MDR algorithm is treated as proprietary information and is not disclosed. In [30] a performance analysis of the MDR algorithm is presented with two trac types: Voice trac with 32 Kbits/sec and 0.0005 arrivals/sec arrival rate and Data trac with 0.1 arrivals/sec arrival rate. Both data and voice trac are modeled with Poisson arrivals. The lengths of the data packets are exponentially distributed with 5120 bytes average length. The voice calls have exponentially distributed durations with a mean of 3 minutes. To the best of our knowledge, the MDR algorithm does not fragment data requests in time and suers from low data throughput as will be discussed shortly.

CHAPTER 5. MEDIUM ACCESS IN THE WIRELESS ATM NETWORK 112 We have simulated our DRAMA algorithm with the input parameters used in the simulation scenario stated in [30]. Our results are shown in Figures 5.33 through 5.36. For voice trac we see that our DRAMA algorithm results in no failures over all of the oered trac range and results in delays that are almost an order of magnitude lower when the oered trac is approximately 80 percent (See Figures 5.33 and 5.34). This result is not surprising since DRAMA gives higher priority to CBR (voice) requests over data requests. As long as the voice requests do not saturate available channel resources DRAMA will not have any CBR failures. The lower and almost constant delay observed in DRAMA versus MDR is also explained by noting the priority level of CBR trac in the DRAMA algorithm. MDR uses a time of expiry based priority scheme that in eect equalizes the access delays observed for both data and voice trac types. For data trac, the results are interesting. In Figure 5.35, we see that when oered trac is higher than 60 percent, DRAMA outperforms MDR. When the offered trac is at 80 percent, DRAMA is an order of magnitude better than MDR. This result is attributed to fragmentation of large data requests across time frames in DRAMA and non-contiguous time sliced allocation in DRAMA. Looking at Figure 5.36, we immediately see the trade-o being made by DRAMA against MDR. MDR has lower average delay when compared to DRAMA but higher failures. The \Last Fragment Delay" as noted in that gure refers to the delay that is experienced by the last fragment of a data packet that was fragmented between time frames. As discussed above, when compared quantitatively, DRAMA is superior to the MDR protocol but requires more complex computational capabilities in the radio port controllers of the wireless ATM network. DRAMA also has the added bene ts of supporting:

Multiple frequencies operating in multiple tiers, Seamless transitions between the tiers in the wireless ATM network, Shared control and signaling formats for unifying the signaling across all tiers of the wireless ATM network.


5.4.2 Polling Multiple Access versus DRAMA

Polling Multiple Access (PMA) is a combination physical layer and link layer protocol proposed in [31]. PMA uses sectored antennas and polling to provide medium access control, resource allocation and to mitigate co-channel interference caused by frequency reuse in wireless ATM local area networks. The transport bit rate proposed in PMA is 160 Mbits/sec with all the radio ports in the local area network sharing the same frequency band. In this case the system capacity is limited by the signal-to-interference (SIR). The authors in [31] report an estimated system capacity of 75 Mbits/sec with an SIR of twenty decibels. The oered trac calculations in PMA are then performed using the interference-adjusted 75 Mbits/sec channel capacity. The resulting interference-adjusted oered trac data in PMA is then equivalent to the oered trac data calculated for DRAMA. The data packet loss probability is reported to vary from 8 10? at an oered trac of 13 percent to 2:5 10? at an oered trac of 93 percent. For continuous bit rate (CBR) trac, the reported call blocking probability is 2 10? up to an oered trac of 66 per cent and climbs to 4 10? at an oered trac of 93 per cent. The delay performance of PMA is not speci ed. PMA is reported to operate best when the oered trac is lower than 75 percent. Unfortunately, [31] does not give any details about the trac models that were being used to evaluate the performance of PMA, hence it was impossible for us to compare directly the performance of DRAMA versus PMA. However, we note that DRAMA has no CBR failures up to an oered trac of 90 percent and its DATA failures are much less than 1 percent in all ranges of oered trac less than 100 percent. DRAMA is a multi-tier protocol that supports both LAN and other environments and does not depend on the physical layer that is being used for medium access control. Based on the comparisons above, DRAMA performs better than PMA in the LAN environment for both CBR and data trac types. 3

2

2

2


5.4.3 DQRUMA versus DRAMA Distributed Queuing Request Update Multiple Access (DQRUMA) is proposed in [29] for providing medium access control and resource allocation in a wireless ATM local area network. DQRUMA provides time of expiry based priority queuing for supporting multiple trac types and uses existing information packets to \piggyback" transmission requests for the next available slots. This results in fewer observed collisions on the signaling channels. DQRUMA uses a round robin packet transmission policy for resource allocation. The details of the trac models used in evaluating the performance of DQRUMA are not disclosed in [29]. The results reported on DQRUMA are based on an in nite timeout period for requests. The access delays seen in DQRUMA are stable up to an oered trac of 85 percent depending on the parameters of the request access algorithm. When the oered trac exceeds 85 percent, the observed delays in DQRUMA grow very rapidly and increase about a factor 10 with an increase of 5 percent in oered trac. These results are obtained with only 1 or 2 packet arrivals per burst. The delays observed in DQRUMA are on the order of 10 time slots up to an oered trac of 85 percent. On a 2 Megabit/sec link 10 time slots yield an average delay of 2 msecs with 53 byte wide time slots. The data trac in DQRUMA is fragmented into 53 byte packets and transmitted. When compared to DQRUMA, DRAMA does not have any failures of CBR trac type up to an oered trac of 90 percent and the data failures are less than 1 percent up to an oered trac of 100 percent. The delay in DRAMA is about 40 msecs up to an oered trac of 90 percent. This delay could be reduced by increasing the transmission rate of the signaling channels. Qualitatively, DRAMA oers the following advantages over DQRUMA:

DRAMA oers full support for multi-tier operation by supporting multiple frequencies and frequency-time sliced resource allocation.

DRAMA may be used even in low bit rate environments without degradation of access delays, where as DQRUMA's 10 time slot average delay degrades with reduced bit rate.


In DQRUMA, even when piggy backing is used, the resources still need to

be allocated at every allocation step. DRAMA reserves resources for CBR trac and eliminates contention for on-going CBR requests. This results in guaranteed bandwidth for CBR requests.

5.5 Conclusion The Dynamic Resource Allocating Multiple Access (DRAMA) algorithm for resource allocation and medium access in a multi-tier wireless ATM network was discussed in this chapter. We rst presented the wireless ATM network environment. Medium Access Control and Resource Allocation in DRAMA were discussed in detail in Sections 5.2.1 and 5.2.2. In Section 5.3, we presented detailed simulations of the DRAMA protocol in wide area, campus wide and local area networks under varied trac assumptions (See Table 5.4 for a summary of our results). Finally, we presented a comparison of DRAMA with other medium access and resource allocation protocols that are proposed in the literature. Based on our analysis, DRAMA makes ecient use of the wireless ATM network resources by oering high success rates for user requests with reasonable access delays. It is the rst medium access control to be proposed for a multi-tier wireless ATM network. Bene ts of DRAMA are summarized below:

DRAMA is a good algorithm for performing medium access control and re-

source allocation in a multi-tier wireless ATM network. It supports multiple users, multiple connections with dierent trac types per user and is designed to function well in a wide-variety of environments. When the failure rate and delay gures for DRAMA are compared to schemes proposed in the literature, DRAMA performs better than or equal to those schemes and has the added bene t of supporting multiple trac types in a multi-tier environment with multiple frequencies.

DRAMA oers acceptable throughput that is close to wireline networks and has acceptable medium access and resource allocation delay. These delays

CHAPTER 5. MEDIUM ACCESS IN THE WIRELESS ATM NETWORK 116 may be reduced either by increasing the bit rate of the signaling channels or by decreasing the signaling frame length in environments that are lightly loaded.

DRAMA scales well between dierent environments. Our results show uniformity in terms of delay and failure rates in both low bit rate and high bit rate environments under a variety of trac conditions. The capacity of the wireless ATM network can be increased by increasing the bandwidth allocated to each frequency of a radio port, thereby increasing the bit rate and the number of slots per frequency, or by increasing the number of frequencies allocated to each radio port. In most cases increasing the bit rates of existing frequencies results in better throughput than increasing the number of frequencies because of the frequency switching problem discussed above.

The collisions observed in the signaling channels are inversely proportional to the signaling bit rate as can be seen in Figure 5.22.

It is possible to support a large number of users in a wireless ATM network with low signaling bit rates as shown by our performance gures.

The failure rates observed in the DRAMA algorithm improve as the carried

trac becomes burstier in nature as observed from the dierences in results between the CWN and LAN environments.

DRAMA is designed to work with Overlay and Migratory Signaling protocols for wireless ATM networks.


Frame N

Frequencies Frequency 1

Frequency 2

Frequency 3

S1

S1

S2

S2

S2

S3

S3

S3

S4

S4

S4

S5

S5

S5

S6

S6

S6

S7

S7

S7

S1

Time Slots

Frequencies

Frame N+1 Frequency 1

Time Slots

Signaling Frequency

Signaling Slots

Signaling Frames User Data Frames

Signaling Frequency

Frequency 2

Frequency 3

S1

S1

S1

S2

S2

S2

S3

S3

S3

S4

S4

S4

S5

S5

S5

S6

S6

S6

S7

S7

S7

Signaling Slots

Figure 5.1: Time Slotted Frequency Multiplexed Radio Link


Channel Matrix Channel Freq 1 Freq 2 I

Time Slots

Channel Chunk Matrix

Freq 3

I

B

B

2,0

I

I

B

1,3

B

I

B

I

I

I

B B

B

I

B

B

B

B

B

B

B

B B

B

B

B

B

B

B

I

B I

B

I

I B

I

B I

B

I B

B B

B: Busy I: Idle

B

Total : 3,-

4,1

1,3

4,-

1,-

Size of Chunk, Position

Figure 5.2: Channel Matrix and Channel Chunk Matrix


Frequency 1

Frequency 2

User A

Time Slots

User A Switching Time

Blind Slot

Frequency Switch

Blind Slot User A User A

Figure 5.3: An Illustration of the Frequency Switching Problem


Connection Request From Higher Layer

Form Request

Synchronize

Send Request

YES Report Failure

Report Success & Connect

Time-out

NO

YES Response

Queued Wait for Response

YES

NO

ACK ?

No Collision, Exponential Backoff

Figure 5.4: Medium Access in the Drama Algorithm


Received Requests

Queued Requests Combine & Prioritize

Save all unsatisfied requests

YES

All Requests Processed ?

NO Get next request

Exact match?

Allocate & Notify Portable

YES

NO Bigger Chunk?

NO

YES

NO Multiple Frequency?

Single Freq. ?

YES

NO

YES

Figure 5.5: Resource Allocation in the Drama Algorithm


Get Next Request

Queued Requests

Allocate, Update Notify

Feasible Find Suitable Frequency

YES

Check Frequency Constraint

NO

NO

Scanned All Frequencies ?

Not Feasible

YES Go to Next Step

Figure 5.6: Flow Diagram for Exact Match and Bigger Chunk Allocation


Determine Request Size

Get Next Request Queued Requests

Request Size

Find Frequency NO

Allocate, Update Notify

Feasible

Check Frequency Constraint

Scanned All Frequencies?

Not Feasible

Go to next step

YES

Figure 5.7: Flow Diagram for Same Frequency Allocation


Get Next Request

Determine Request Size

Save Request into queue YES NO

Scanned All Frequencies ?

Find least occupied frequency NO

Allocate as many slots as possible in this frequency and decrement request size by the number of allocated slots. Allocate, Notify Update

New Request Size ?= 0

YES

Figure 5.8: Flow Diagram for Multiple Frequency Allocation


80

70

60

50

40

30

20

10

0 0

100

200

300

400

500

600

700

800

900

1000

Figure 5.9: Time Series Drawn from Pareto Distribution (Mean = 1)


80

70

60

50

40

30

20

10

0 0

100

200

300

400

500

600

700

800

900

1000

Figure 5.10: Time Series Drawn from Exponential Distribution (Mean = 1)


0.4 0.35

Failure Rate

0.3 0.25 0.2 0.15 0.1 CBR Failure Rate Data Failure Rate 0.05 0 0

0.2

0.4

0.6

0.8

1

1.2

Normalized Offered Traffic

Figure 5.11: Wide Area Network Failure Rates (Poisson CBR and Data Arrivals)


0.5 0.45 0.4

Seconds

0.35 0.3 0.25 0.2 0.15

CBR Access Delay Data Access Delay

0.1 0.05 0 0

0.2

0.4

0.6

0.8

1

1.2


Figure 5.12: Wide Area Network Average Access Delays (Poisson CBR and Data Arrivals)


0.9 0.8

4 Kbps 8 Kbps

WAN Fragmentation Ratio

0.7

16 Kbps

0.6 0.5 0.4 0.3 0.2 0.1 0 0

0.2

0.4 0.6 0.8 Offered Traffic (Voice+Data)

1

1.2

Figure 5.13: Wide Area Network Data Fragmentation Ratio (Poisson CBR and Data Arrivals)


0.04 16 Kbps 32 Kbps

Collisions on the Signaling Channels

0.035

0.03

0.025

0.02

0.015

0.01

0.005

0 0

0.1

0.2

0.3

0.4 0.5 0.6 0.7 Offered Traffic (CBR + Data)

0.8

0.9

1

Figure 5.14: Collision Rates with 16 and 32 Kbits/sec signaling in the Campus Wide Network (Poisson CBR and Self Similar Data Trac Arrivals)


0.03 95% Confidence Interval Lower Bound 95% Confidence Interval Upper Bound 16 Kbps 32 Kbps

CBR Failure Rate

0.02

0.01

0

0.1

0.2

0.3

0.4 0.5 0.6 0.7 Offered Traffic (CBR+Data)

0.8

0.9

Figure 5.15: CBR Failures with 16 and 32 Kbits/sec signaling in the Campus Wide Network (Poisson CBR and Self Similar Data Trac Arrivals)


0.025 95% Confidence Interval Lower Bound 0.02

95% Confidence Interval Upper Bound 16 Kbps 32 Kbps

Data Failure Rate

0.015

0.01

0.005

0

−0.005 0.1

0.2

0.3


0.8

0.9

Figure 5.16: Data Failures with 16 and 32 Kbits/sec signaling in the Campus Wide Network (Poisson CBR and Self Similar Data Trac Arrivals)



CBR Access Delay (Seconds)

0.05

0.04

0.03

0.02

0.01

0 0

0.1

0.2

0.3


0.8

0.9

1

Figure 5.17: CBR Access Delay with 16 and 32 Kbits/sec signaling in the Campus Wide Network (Poisson CBR and Self Similar Data Trac Arrivals)



0.08

Data Access Delay (Seconds)

0.07 0.06 0.05 0.04 0.03 0.02 0.01 0 0.1

0.2

0.3


0.8

0.9

1

Figure 5.18: Data Access Delay with 16 and 32 Kbits/sec signaling in the Campus Wide Network (Poisson CBR and Self Similar Data Trac Arrivals)


0.2 0.18

16 Kbps 32 Kbps

Last Fragment Delay (Seconds)

0.16 0.14 0.12 0.1 0.08 0.06 0.04 0.02 0 0.2

0.3

0.4


0.9

1

Figure 5.19: Last Fragment Delays with 16 and 32 Kbits/sec signaling in the Campus Wide Network (Poisson CBR and Self Similar Data Trac Arrivals)


0.8

0.7

16 Kbps 32 Kbps

Data Fragmentation Ratio

0.6

0.5

0.4

0.3

0.2

0.1

0 0

0.5 1 Offered Traffic (Voice + Data)

1.5

Figure 5.20: Data Fragmentation Ratio with 16 and 32 Kbits/sec signaling in the Campus Wide Network (Poisson CBR and Self Similar Data Trac Arrivals)


0.35

0.3

Failure Rate

0.25

0.2

0.15

0.1 CBR Failure Rate Data Failure Rate 0.05

0 0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6


Figure 5.21: Campus Wide Network Failure Rates with Poisson CBR and Data Arrivals



Collision on the Signaling Channels

0.14

0.12

0.1

0.08

0.06

0.04

0.02

0 0.1

0.2

0.3


0.8

0.9

1

Figure 5.22: Collisions on the Signaling Channel with 32 and 64 Kbits/sec Signaling Bit Rates in the Local Area Network (Poisson CBR and Self Similar Data Trac Arrivals)


−3

5

x 10

95% Confidence Interval Lower Bound 4

95% Confidence Interval Upper Bound 32 Kbps

CBR Failure Rate

3

64 Kbps

2

1

0

−1

−2 0.2

0.3

0.4

0.5 0.6 0.7 Offered Traffic (CBR+Data)

0.8

0.9

Figure 5.23: CBR Failure Rates with 32 and 64 Kbits/sec Signaling Bit Rates in the Local Area Network (Poisson CBR and Self Similar Data Trac Arrivals)


0.02 95% Confidence Interval Lower Bound 0.015

95% Confidence Interval Upper Bound 32 Kbps 64 Kbps

Data Failure Rate

0.01

0.005

0

−0.005

−0.01 0.2

0.3

0.4

0.5 0.6 0.7 Offered Traffic (CBR+Data)

0.8

0.9

Figure 5.24: Data Failure Rates with 32 and 64 Kbits/sec Signaling Bit Rates in the Local Area Network (Poisson CBR and Self Similar Data Trac Arrivals)


0.1 32 Kbps 64 Kbps

0.09


0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0 0.2

0.3

0.4

0.5 0.6 0.7 Offered Traffic (CBR + Data)

0.8

0.9

1

Figure 5.25: CBR Access Delay with 32 and 64 Kbits/sec Signaling Bit Rates in the Local Area Network (Poisson CBR and Self Similar Data Trac Arrivals)


0.045 32 Kbps 64 Kbps


0.04

0.035

0.03

0.025

0.02

0.015 0.1

0.2

0.3


0.8

0.9

1

Figure 5.26: Data Access Delay with 32 and 64 Kbits/sec Signaling Bit Rates in the Local Area Network (Poisson CBR and Self Similar Data Trac Arrivals)



0.3

Fragmentation Ratio

0.25

0.2

0.15

0.1

0.05

0 0.1

0.2

0.3


0.8

0.9

1

Figure 5.27: Fragmentation Ratio with 32 and 64 Kbits/sec Signaling Bit Rates in the Local Area Network (Poisson CBR and Self Similar Data Trac Arrivals)



Last Fragment Access Delay (Seconds)

0.105 0.1 0.095 0.09 0.085 0.08 0.075 0.07 0.065 0.06 0.1

0.2

0.3


0.8

0.9

1

Figure 5.28: Last Fragment Delay with 32 and 64 Kbits/sec Signaling Bit Rates in the Local Area Network (Poisson CBR and Self Similar Data Trac Arrivals)


0.2 0.18

Actual Value Lower Bound

0.16

Upper Bound

0.14

Seconds

0.12 0.1 0.08 0.06 0.04 0.02 0 0

0.5 1 Offered Traffic ( CBR + Data)

1.5

Figure 5.29: 95% Con dence Intervals for CBR Access Delay using 32 Kbits/sec signaling and Poisson CBR and Self Similar Data Trac Arrivals in the Campus Wide Network (See Figure 5.17)


0.09 Actual Value Lower Bound Upper Bound

0.08


0.07 0.06 0.05 0.04 0.03 0.02 0.01 0 0.1

0.2

0.3

0.4


0.9

1

1.1

Figure 5.30: 95% Con dence Intervals for Data Access Delay using 32 Kbits/sec signaling in the Campus Wide Network for Poisson CBR and Self Similar Data Trac Arrivals (See Figure 5.18)


0.02 0.018

Actual Value Lower Bound


0.016

Upper Bound

0.014 0.012 0.01 0.008 0.006 0.004 0.002 0 0.2

0.3

0.4

0.5 0.6 0.7 Offered Traffic (CBR + Data)

0.8

0.9

1

Figure 5.31: 95% Con dence Intervals for CBR Access Delay using 64 Kbits/sec signaling and Poisson CBR and Self Similar Data Trac Arrivals in the Local Area Network (See Figure 5.25)


0.055

0.05

Actual Value Lower Bound Upper Bound


0.045

0.04

0.035

0.03

0.025

0.02

0.015 0.1

0.2

0.3


0.8

0.9

1

Figure 5.32: 95% Con dence Intervals for Data Access Delay using 64 Kbits/sec signaling in the Local Area Network for Poisson CBR and Self Similar Data Trac Arrivals (See Figure 5.26)


0.1

0.08

MDR Voice Failure Rate DRAMA Voice Failure Rate

Failure Rate

0.06

0.04

0.02

0

−0.02 0.1

0.2

0.3

0.4 0.5 0.6 Offered Traffic (Voice + Data)

0.7

0.8

Figure 5.33: Comparison of Voice Failure Rates (Poisson Data and Voice Arrivals)


0.2 0.18

MDR Voice Access Delay DRAMA Voice Access Delay

0.16

Access Delay (Sec)

0.14 0.12 0.1 0.08 0.06 0.04 0.02 0 0.2

0.3


0.7

0.8

Figure 5.34: Comparison of Voice Access Delays (Poisson Data and Voice Arrivals)


0.01 0.009

MDR DATA Failure Rate DRAMA DATA Failure Rate

0.008

Failure Rate

0.007 0.006 0.005 0.004 0.003 0.002 0.001 0 0.2

0.3


0.7

0.8

Figure 5.35: Comparison of Data Failure Rates (Poisson Data and Voice Arrivals)


0.25 MDR DATA Access Delay

Access Delay (Sec)

0.2

DRAMA DATA Access Delay DRAMA Last Fragment Access Delay

0.15

0.1

0.05

0 0.2

0.3


0.7

0.8

Figure 5.36: Comparison of Data Access Delays (Poisson Data and Voice Arrivals)


Criteria CBR Rate

LAN

Failure No CBR failures up to 90 percent oered traf c and less than 1 percent up to 100 percent offered trac (150 users supported by 3 channels of 10 Mbits/sec) Data Failure Less than 1 perRate cent up to 90 percent oered trac and below 3 percent up to 100 percent offered trac Average CBR Less Access Delay than 20 msecs up to 80 percent oered trac Average Data Less Access Delay than 45 msecs up to 85 percent oered trac Collision Less than 7 percent up to an oered trac of 100 percent

Campus

WAN

Less than 1 percent up to 90 percent oered trac

Less than 4.1 percent up to 83 percent oered trac

Less than 10 msecs up to 80 percent oered trac Less than 10 msecs up to 80 percent oered trac Less than 20 percent up to an oered trac of 100 percent

Less than 40 msecs up to 80 percent oered trac Less than 100 msecs up to an oered trac of 80 percent Less than 1 percent up to an oered trac of 100 percent

Less than 1 percent up to 88 percent offered trac (50 users supported by 5 channels of 2 Mbits/sec)

Table 5.4: Summary of Results

Less than 1 percent up to 83 percent oered trac (375 users supported by 10 channels of 256 Kbits/sec)

Chapter 6 Conclusion Section 6.1 summarizes the main results of this dissertation. Section 6.2 outlines areas for future research.

6.1 Summary of Results This dissertation proposes a possible architecture for a future wireless ATM network that is capable of supporting multiple trac types with dierent quality of service requirements and priorities. Chapter 2 describes our wireless ATM network architecture and explains the terminology that is used in this dissertation. Chapter 3 describes on two alternatives for implementing signaling in the wireless ATM network to support wireless users. Overlay Signaling is designed to support mobility using the existing ATM protocols. It implements this support using a signaling network overlaid on the backbone ATM network using the Zone concept described in Chapter 2. Overlay Signaling is the only wireless ATM network signaling protocol proposed in the literature that is able to function using the current ATM signaling protocols. On the other hand, Migratory Signaling enhances the existing ATM signaling protocols by adding signaling messages to support wireless users natively. When compared to Overlay Signaling, Migratory Signaling uses less bandwidth for signaling and does not need additional signaling circuits that are dedicated for support of wireless users. Overlay Signaling implements the support for 154

CHAPTER 6. CONCLUSION

155

wireline and wireless user interaction by using service gateways. Migratory Signaling provides built-in support for wireless users. Chapter 4 focuses on solving the problem of dynamic rerouting of user connections during a hando event. We present a rerouting method based on cell forwarding for wireless ATM networks that use Overlay Signaling. Although cell forwarding has some disadvantages as described later in the chapter, it is very suitable for the task of supporting wireless users in a wireless ATM network that uses Overlay Signaling. The Nearest Common Node Rerouting (NCNR) is a protocol for dynamic rerouting of user connections in wireless ATM networks utilizing Migratory Signaling. NCNR is based on rerouting a user connection at the node that is the closest common ancestor of the zones that are involved in the hando. NCNR makes maximum re-use of the existing user connection during rerouting and saves network resources. When compared to alternate rerouting schemes previously proposed in the literature, NCNR consistently involves fewer network nodes and requires fewer signaling messages for rerouting. Also in Chapter 4, we identify the dierent needs of time sensitive and throughput dependent trac types during a hando and showed how hando can be implemented to better suit the needs of these types of trac. Chapter 5 focuses on the medium access control and resource allocation aspects of wireless ATM networks. Dynamic Resource Allocating Multiple Access (DRAMA) is devised to perform medium access control and resource allocation in a multi-tier wireless ATM network. DRAMA is a reservation based medium access control that is unique in its support for multiple frequency bands and frequency-time sliced resource allocation. We present detailed simulations of the operation of DRAMA and analyzed the performance of the protocol under a variety of operating conditions. In high bit rate campus wide and local area network environments, resource allocation success rates of over 90 percent are shown to be feasible in our simulations. The results of our simulations indicate that DRAMA performs better than or equal to previously proposed medium access control protocols for wireless ATM networks while providing the added bene ts discussed in Chapter 5. While evaluating the performance of DRAMA, we develop baseline trac models for wireless ATM networks and simulation techniques that are suitable for simulating self similar trac


156

patterns.

6.2 Topics for Future Research The following is a list of issues for future research:

Resource Allocation in Dynamic Resource Allocating Multiple Access: In Chap-

ter 5, we have developed the DRAMA algorithm using a heuristic and greedy resource allocation strategy. As stated in that chapter, it may be desirable to develop a resource allocation policy that considers the existing requests, the requests that are waiting to be allocated and the frequency switching constraints, and re-arranges the channel matrix to allocate the highest possible number of requests. The speed and the success of the resource allocation algorithm may then be compared to the heuristic resource allocation utilized in DRAMA.

Trac Models for Wireless ATM Networks: Although we have discussed possi-

ble trac models for wireless ATM networks in Chapter 5, as wireless networks become more ubiquitous and as support for higher bit rates is implemented, trac models based on operating wireless networks need to be constructed. Using these trac models and mobility models introduced in [9], a worst-case analysis of the number of signaling transactions that may occur in a wireless ATM network may be developed to complement the analysis presented in Chapter 3.

Co-Channel Interference Mitigation During Resource Allocation: In Chapter

5, we have brie y mentioned that co-channel interference may be mitigated by dynamic channel allocation or by interference cancellation techniques. Techniques for combining resource allocation with dynamic channel allocation in order to mitigate co-channel interference in frequency reuse based systems need to be developed in future research.


157

Database Support for Mobility in Wireless ATM Networks: We have assumed

an HLR/VLR based database architecture for supporting mobility in the wireless ATM network in this dissertation. Hierarchical and replicated database architectures require less signaling and less time to locate a user. Although database architectures for existing cellular phone networks are actively being researched, the applicability of these techniques to wireless ATM networks needs to be investigated.

Adaptation of User Applications to Multi-tier Networks: In Chapter 5, we

discussed the topic of multi-tier wireless ATM networks. DRAMA provides support for seamless transition between the tiers in the wireless ATM network. However, the user applications need to adapt to varying bandwidth conditions and resource availability. The issue of adapting applications to dierent resource availability needs to be investigated. An example of such an application is Pyramidal Coding designed for image transmission across heterogeneous networks[51].

Appendix A ATM Signaling Message Lengths The ATM Signaling messages mentioned in the text are taken from [16]. According to the speci cations, the lengths of ATM signaling messages may vary depending on the situation. We will list the minimum and maximum lengths for these messages in Table A.1. Table A.1: ATM Signaling Message Lengths ATM Signaling Message

Minimum Length (bytes) Maximum Length (bytes) SETUP 107 238 CALL PROCEEDING 17 25 CONNECT 25 53 CONNECT ACK. 9 9 RELEASE 15 43 RELEASE COMPLETE 13 43

158

Appendix B Derivation of Equations 3.1 through 3.3 In [19], equations for estimating the mean of database transaction rates are given. We adapted these equations to the wireless ATM network and simpli ed them. The derivation of these equations are given below:

The rst equation Lc = Npc(1 + q) stays the same in both texts. The equation given for registration in [19] is: ? m) (evP Mj + N v4D(1 ? m )): Lr = (1 ? s)(1 p r 2

The dierences between [19] and this text are:

{ Nr is equal to 1 since the call processing is fully distributed. The value of m is assumed to be zero for worst case analysis. stands for the population density in a zone. 2

{ Registration transaction does not include the hando location updates

which are accounted for in the hando transaction, so we omit the evPpMj part of the equation resulting in ?s ?m Nr v4D. Multiply and divide this equation by D and group D as N , the number of people in a zone, (1 2

159

)(1

)

APPENDIX B. DERIVATION OF EQUATIONS 3.1 THROUGH 3.3 and de ne T = Dv . After these modi cations we obtain at given in the text.

The equation given in [19] for hando is Lh =

?s)(1?m) evP

160

N (1?s)(1?m) T

4

p Mj .

as

The product MjPp simpli es to 4D. Then if we multiply and divide by D and use T given above we arrive at ?s ?m e4 NT . Note that Pp stands for the perimeter of a radio port coverage area and d = Pp stands for one side of a zone assuming a square zone con guration. (1

)(1

)

4

(1

Appendix C Migratory Signaling Messages In this appendix we will specify the contents and the lengths in octets of the proposed W-ATM signaling messages. We follow the format of the ATM User Network Interface Speci cation for the ATM Signaling Messages. We also assume that all of the identi cation numbers and the ATM addresses mentioned below are in E.164 addressing format as de ned by the ITU[6].

C.1 REGISTER The REGISTER message is used to initiate a location update as part of the registration process. It contains the IDN for the user, the ATM address of the present zone and the ATM address of the permanent home database for the user. The elds in this message are listed in Table C.1. 1

C.2 REGISTER COMPLETE The REGISTER COMPLETE message is sent to denote the successful completion of a registration request. It contains the user IDN, the user pro le, the ATM address of the current zone the user is in and the ATM address of the permanent 1

The ATM addresses are limited to be 25 bytes by the ITU speci cation Q.2931.

161

APPENDIX C. MIGRATORY SIGNALING MESSAGES

162

home database for the user. The elds of the REGISTER COMPLETE message are listed in Table C.1.

C.3 RECORD DELETE The RECORD DELETE message is sent from the permanent home database to the previous zone of the user. It contains the user IDN, the ATM address of the permanent home database and the ATM address of the previous zone of the user. The elds of the RECORD DELETE message are listed in Table C.1.

C.4 REGISTER DENY The REGISTER DENY message is sent from the home database to the zone that requests the registration if the user is not authenticated. It contains the user IDN, the ATM address of the permanent home database and the ATM address of the current zone of the user. The elds of the REGISTER DENY message are listed in Table C.1.

C.5 HANDOFF INIT The HANDOFF INIT message is used to initiate a hando. Depending on the algorithm being used it may be sent from the candidate zone to the user's previous zone or it may be sent in the opposite direction. This message contains the user pro le when it is sent from the previous zone to the candidate zone. The elds of the HANDOFF INIT message are listed in Table C.2.

C.6 HANDOFF INFO The HANDOFF INFO message is sent from the user's previous zone to the candidate zone when the hando is being performed through the candidate zone. It


163

Information Element Length (bytes) Protocol Discriminator 1 Message Type 2 Message Length 2 User IDN 8 Originator ATM Address 25 Destination ATM Address 25 User Pro le 1024 Only for REGISTER COMPLETE Total 63

Table C.1: Message contents for registration related messages.

Information Element Length (bytes) Protocol Discriminator 1 Message Type 2 Message Length 2 User IDN 8 Originator ATM Address 25 Destination ATM Address 25 Hando Type Indicator 1 User Pro le 1024 Total 1088

Table C.2: Contents of the HANDOFF INIT message. The user pro le part of the message is optional depending on the hando type indicator.


164

contains the user IDN and the user pro le. The contents of the message are listed in Table C.3.

C.7 HANDOFF CHANNEL ASSIGN The HANDOFF CHANNEL ASSIGN message is sent from the candidate zone to the user's previous zone. It contains the ATM address of the candidate zone, the ATM address of the previous zone, the user IDN and the channel assignment information. The contents of this message are listed in Table C.4.

C.8 HANDOFF REROUTE The HANDOFF REROUTE message is sent from the user's previous zone to the end point for the user connection. It contains the user IDN, the ATM address of the candidate zone, the ATM address of the previous zone and the ATM address of the end point for the user connection. The contents of this message are listed in Table C.5.

C.9 REROUTE CONNECTION The REROUTE CONNECTION message is sent from the end point for the user connection to the candidate zone to perform the rerouting necessary for hando. It contains the user IDN, the ATM address of the candidate zone, the ATM address of the previous zone and the ATM address of the end point. The contents of the REROUTE CONNECTION message are listed in Table C.5.

C.10 REROUTE COMPLETE The REROUTE COMPLETE message is sent from the candidate zone to the end point. It contains the user IDN, the ATM address of the candidate zone and the


Information Element Length (bytes) Protocol Discriminator 1 Message Type 2 Message Length 2 User IDN 8 Originator ATM Address 25 Destination ATM Address 25 User Pro le 1024 Total 1087

Table C.3: Contents of the HANDOFF INFO message.

Information Element Length (bytes) Protocol Discriminator 1 Message Type 2 Message Length 2 User IDN 8 Originator ATM Address 25 Destination ATM Address 25 Channel Assignment Information 4 Total 67

Table C.4: Contents of the HANDOFF CHANNEL ASSIGN message.

165


166

ATM address of the end point. The contents of this message are listed in Table C.6.

C.11 HANDOFF COMPLETE The HANDOFF COMPLETE message is sent from the candidate zone to both the previous zone and the end point for the user connection. It contains the user IDN, the ATM address of the candidate zone, the ATM address of the end point and the ATM address of the previous zone. The contents of the HANDOFF COMPLETE message follow the format given in Table C.5.

C.12 HANDOFF FAIL The HANDOFF FAIL message is sent from the candidate zone to the previous zone and to the end point for the user connection. It contains the user IDN, the ATM address of the candidate zone, the ATM address of the end point and the ATM address of the previous zone. The contents of this message are listed in Table C.5.

C.13 RECORD UPDATE The RECORD UPDATE message is sent from the candidate zone to the permanent home database of the user to update the user records. It contains the user IDN, the ATM address of the candidate zone and the ATM address of the permanent home database. This message follows the format given in Table C.1.

C.14 LOCATION REQUEST The LOCATION REQUEST message is sent from the calling party's zone to the home database of the called party. If the calling party is a xed ATM host then the message is sent directly from the calling party to the home database of the called party. It contains the calling party's IDN, the called party's IDN, the ATM address


167

Information Element Length (bytes) Protocol Discriminator 1 Call Reference 4 Message Type 2 Message Length 2 User IDN 8 Endpoint ATM Address 25 Previous Zone ATM Address 25 Candidate Zone ATM address 25 Total 92

Table C.5: Contents of HANDOFF REROUTE, REROUTE CONNECTION , HANDOFF COMPLETE and HANDOFF FAIL messages.

Information Element Length (bytes) Protocol Discriminator 1 Call Reference 4 Message Type 2 Message Length 2 User IDN 8 Endpoint ATM Address 25 Candidate Zone ATM address 25 Total 67

Table C.6: Contents of the REROUTE COMPLETE message.


168

of the calling party's zone and the ATM address of the home database of the called party. If the calling party is a xed ATM host then the user IDN is set to the E.164 part of the ATM address of the calling party. The contents of the message are listed in Table C.7.

C.15 LOCATION REPLY The LOCATION REPLY message is sent from the home database for the called party to the calling party or the zone of the calling party. It contains the user IDN for the called party, the user IDN for the calling party, the ATM address of the called party's location, the ATM address of the permanent home database and the ATM address of the calling party or the zone of the calling party. The contents of the message are listed in Table C.8.

C.16 REQUEST DENIED The REQUEST DENIED message is sent from the home database to the calling party or the zone of the calling party. It contains the user IDN of the called party, the ATM address of the permanent home database, the ATM address of the calling party or the zone of the calling party and the reason for refusal. The contents of this message are listed in Table C.9.

C.17 MOBILE SETUP The MOBILE SETUP message is sent from the calling party to called party. The contents of this message vary depending on which party is a mobile or a wireless user. In general it contains the user IDN's for both parties, the ATM addresses for both parties and the connection parameters. This message follows the format of ATM UNI SETUP message very closely. See Section 3.2.3 for details. The contents of the MOBILE SETUP message is listed in Table C.10.

APPENDIX C. MIGRATORY SIGNALING MESSAGES Information Element Length (bytes) Protocol Discriminator 1 Message Type 2 Message Length 2 User IDN 8 Calling Party ATM Address 25 Called Party ATM address 25 Total 63

Table C.7: Contents of the LOCATION REQUEST message. Information Element Length (bytes) Protocol Discriminator 1 Message Type 2 Message Length 2 Calling Party User IDN 8 Called Party User IDN 8 Calling Party ATM Address 25 Called Party ATM address 25 Location for Called Party 25 Total 96

Table C.8: Contents of the LOCATION REPLY message. Information Element Length (bytes) Protocol Discriminator 1 Message Type 2 Message Length 2 User IDN 8 Calling Party ATM Address 25 Called Party ATM address 25 Cause 1 Total 64

Table C.9: Contents of the REQUEST DENIED message.

169


170

Information Element Length (bytes) Protocol Discriminator 1 Call Reference 4 Message Type 2 Message Length 2 Connection Type 1 AAL Parameters 4-21 ATM Trac Descriptor 12-30 Broadband Bearer Capability 6-7 Calling User IDN 8 Called User IDN 8 Calling Party ATM Address 25 Called Party ATM address 25 Connection Identi er 9 Quality of Service (QoS) Parameter 6 Endpoint Reference 4-7 Total 117-156

Table C.10: Contents of the MOBILE SETUP message. Note that some of the elds are included directly from the ATM UNI SETUP signaling message. Optional elds have been omitted.


171

C.18 MOBILE SETUP DENIED The MOBILE SETUP DENIED message is sent when a MOBILE SETUP message is not honored by an ATM node. It contains the cause of refusal, the ATM addresses of the end points in the connection and the ATM address of the node that originates the message. The contents of the message are listed in Table C.11.

C.19 MOBILE ACCEPT The MOBILE ACCEPT message is sent from the called party to the calling party . It contains the IDN's for both the called and calling parties, the ATM addresses of the two end points. The contents of this message are listed in Table C.12. 2

C.20 MOBILE START The MOBILE START message is sent in response to the MOBILE ACCEPT message and its contents are also listed in Table C.12.

C.21 MOBILE CLEAR The MOBILE CLEAR message is sent to clear an existing connection. It contains the ATM addresses of the two end points for the connection and the user IDN's for both of the users. The contents of the MOBILE CLEAR message are listed in Table C.12.

C.22 MOBILE CLEAR ACK The MOBILE CLEAR ACK message is sent in response to the MOBILE CLEAR message and its contents are also listed in Table C.12. 2

This could also be the zone managers for either of the parties. See Section 3.2.3 for details.


Information Element Length (bytes) Protocol Discriminator 1 Call Reference 4 Message Type 2 Message Length 2 Connection Type 1 Calling User IDN 8 Called User IDN 8 Calling Party ATM Address 25 Called Party ATM address 25 Originator ATM Address 25 Connection Identi er 9 Cause of Refusal 2 Total 112

Table C.11: Contents of the MOBILE SETUP DENIED message.

Information Element Length (bytes) Protocol Discriminator 1 Call Reference 4 Message Type 2 Message Length 2 Connection Type 1 Calling User IDN 8 Called User IDN 8 Calling Party ATM Address 25 Called Party ATM address 25 Connection Identi er 9 Total 85

Table C.12: Contents of the signaling messages related to Connection Setup.

172

Bibliography [1] Graph of NAP/TREND throughput for May 19, 1997, Paci c Bell, http://www.pacbell.com/products/business/fastrak/networking/nap/index.html. [2] QuickCam Product Literature, Connectix Corporation, San Mateo, CA., USA. [3] Fore Systems ASX-200BX and ASX-1000 product literature, Fore Systems, Warrendale, PA., USA. [4] D. C. Cox, \Wireless Personal Communications: What is it?," IEEE Personal Communications Magazine, Vol. 2, No. 2, April 1995, pp. 2-35. [5] Siegmund H. Redl, M. K. Weber and M. W. Oliphant, An Introduction to GSM, Boston: Artech House, 1995. [6] Q.2931 ATM Network Signaling Speci cation, ITU. [7] B. Rajagopalan, \Mobility Management in Integrated Wireless ATM Networks," Mobicom' 95, November 1995, Berkeley, CA., USA. [8] R. Jain, S. Rajagopalan, L. F. Chang, \Phone Number Portability for PCS systems with ATM Backbones Using Distributed Dynamic Hashing," submitted to IEEE JSAC. [9] J. Jannink, D. Lam, N. Shivakumar, J. Widom, and D. C. Cox, \Ecient and Flexible Location Management Techniques for Wireless Communications Systems," Submitted for journal publication. 173

BIBLIOGRAPHY

174

[10] D. C. Cox, \A Radio System Proposal for Widespread Low-Power Tetherless Communications," IEEE Trans. on Communications, Vol. 39, No. 2, pp. 324335, Feb. 1991. [11] Y. Lee and D. C. Cox, MAP Selection Diversity DFE for Indoor Wireless Data Communications," IEEE ICUPC '96 Digest, Cambridge, MA, September 29 October 2, 1996. [12] T. M. Schmidl and D. C. Cox, \Low-overhead, Low-complexity Burst Synchronization for OFDM," IEEE ICC '96 Digest, Dallas, TX, June 23-27, 1996. [13] B-J. Kim and D. C. Cox, \A Blind Equalization Method for Short Burst Communications over Frequency Selective Wireless Channels," submitted to IEEE Trans. on Comm. [14] D. C. Cox, \Wireless Network Access for Personal Communications," IEEE Communications Magazine, Vol. 30, No. 12, pp. 96-115, Dec. 1992. [15] EIA/TIA interim standard 41. Cellular radio telecommunications intersystem operations., Washington, DC. : Electronic Industries Association, 1991. [16] ATM Forum User Network Interface speci cation version 3.1, ATM Forum, May 1994. [17] A. S. Acampora, \An architecture and methodology for mobile-executed hando in cellular ATM networks," IEEE Journal on Selected Areas in Communications, Oct. 1994, vol.12, no.8, p. 1365-75. [18] D. Raychaudhuri et al., \ATM Based Transport Architecture for Multiservices Wireless Personal Communication Networks," IEEE J. on Selected Areas in Comm., October 1994, pp 1401-1414. [19] C. N. Lo et al., \An Estimate of Network Database Transaction Volume to Support Universal PCS," ICUPC '92, Dallas, TX., 29 Sept.-1 Oct. 1992.

BIBLIOGRAPHY

175

[20] LATA Switching Systems Generic Requirements, Bellcore Technical Reference TR-TSY-000517, Issue 3, March 1989, Sec. 17, Table 17.6-B:\Trac Capacity and Environment." [21] R. Yuan, S. Biswas et al., \Mobility Support in a Wireless ATM Network," 5th Workshop on Third Generation Wireless Information Networks, 1995, Kluwer Publishers, pp. 335-345. [22] K. Y. Eng et al., \A Wireless Broadband ad-hoc ATM local-area Network," Wireless Networks, vol. 1 , pp. 161-174, 1995. [23] O. T. W. Yu and V. C. M. Leung, \B-ISDN Architectures and Protocols to Support Wireless Personal Communications Internetworking," PIMRC' 95, Sep. 27-29, 1995, Toronto, Canada. [24] C. Toh, \The Design and Implementation of a Hybrid Handover Protocol for Multimedia Wireless LANs," Mobicom 95, Nov. 13-15, 1995, Berkeley, CA., U.S.A. [25] C. Hedrick, \Routing Information Protocol", RFC 1058, Rutgers University, June 1988. [26] L. Kleinrock and F. Tobagi, \Packet Switching in Radio Channels: Part ICarrier Sense Multiple Access Modes and Their Throughput Delay Characteristics," IEEE Trans. on Communications, Vol. COM-23, pp. 1400-1416, Dec. 1975. [27] L. Kleinrock and F. Tobagi, \Packet Switching in Radio Channels: Part II-The Hidden Terminal Problem in Carrier Sense Multiple Access and the Busy Tone Solution," IEEE Trans on Communications, Vol. COM-23, pp. 1417-1433, Dec. 1975. [28] L. Kleinrock and F. Tobagi, \Packet Switching in Radio Channels: Part IIIPolling and Dynamic Split Channel Reservation Multiple Access," IEEE Trans on Communications, Vol. COM-24, pp. 832-845, Aug. 1976.

BIBLIOGRAPHY

176

[29] M. J. Karol, Z. Liu, K. Y. Eng, \Distributed Queueing Request Update Multiple Access (DQRUMA) for wireless packet (ATM) networks," Proceedings of the IEEE ICC 95 , pp. 1224-1231. [30] D. Raychaudhuri, N. Wilson, \ATM-based transport architectures for multiservices wireless personal communication networks," IEEE Journal on Selected Areas in Communications, Vol. 12, Oct 1994. p 1401-1414. [31] A. Mahmoud, D. Falconer, S. Mahmoud, \Multiple access scheme for wireless access to a broadband ATM LAN based on polling and sectored antennas," Proceedings of the IEEE PIMRC 95, Toronto, Canada, pp. 1047-1051. [32] E. Ayanoglu, K. Y. Eng and M. J. Karol, \Wireless ATM: Limits, Challenges and Proposals," IEEE Personal Communications Magazine, Vol. 3. No. 4, August 1996, pp. 18-35. [33] R. R. Gejji, \Mobile Multimedia Scenario Using ATM and Microcellular Technologies," IEEE Trans. on Veh. Tech., Vol. 43, No. 3, pp. 699-703, Aug. 1994. [34] Joseph F. Rizzo and Nelson R. Sollenberger, \Multitier Wireless Access," IEEE Personal Communications Magazine, June 1995. [35] D. Raychaudhuri, \Wireless ATM Networks: Architecture, System Design and Prototyping," IEEE Personal Communications Magazine, Vol. 3. No. 4, August 1996, pp. 42-49. [36] A. Ganz, Z. Haas, C. Krishna, \Multi-tier Wireless Networks for PCS," Proceedings of the 1996 IEEE 46th Vehicular Technology Conference, Atlanta, GA, USA, pp. 436-440. [37] M. J. Karol, Z. J. Haas and C. B. Woodworth, \Performance Advantages of Time-Frequency-Sliced Systems," Proceedings of the IEEE PIMRC' 95, Toronto, Canada, pp. 1104-1111.

BIBLIOGRAPHY

177

[38] J. C. Chuang, D. C. Cox and N. R. Sollenberger, \Pilot-based dynamic channel assignment scheme for wireless access TDMA/FDMA systems," Int. J. of Wireless Information Networks, Vol. 1, No. 1, Jan 1994, pp. 37-47. [39] R. Baines, \The Soft Approach," Telephony, Vol. 227, No. 4, July 1994. [40] E. Kooistra, \A software implemented spread spectrum modem based on two TMS320C50 DSPs," Proceedings of the Third Symposium on Communications and Vehicular Technology, Eindhoven, NL, 25-26 October 1995, pp. 91-96. [41] D-G. Jeong, W-S. Jeon, \Performance of an Exponential Backo scheme for slotted-ALOHA protocol in local wireless environment," IEEE Trans. on Vehicular Technology, Vol. 4, No. 3, August 1995, pp. 470-479. [42] M. Cheng, J. C. Chuang, \Distributed measurement-based quasi- xed frequency assignment for personal communications," Proceedings of the ICC 95, Seattle, WA, USA, pp. 433-437 [43] Microwave Mobile Communications, William C. Jakes, New York, Wiley, 1974. [44] W. Willinger, M. Taqqu, R. Sherman and Daniel V. Wilson, \Self Similarity Through High Variability: Statistical Analysis of Ethernet LAN Trac at the Source Level," IEEE/ACM Trans. on Networking, February 1997. [45] V. Paxson, \Empirically Derived Analytic Models of Wide Area TCP Connection," IEEE/ACM Trans. on Networking, Vol. 2, No. 4, August 1994, pp. 316-336. [46] V. Paxson, S. Floyd, \Wide Area Trac: The failure of Poisson Modeling," IEEE/ACM Trans. on Networking, Vol. 3, No. 3, June 1995, pp. 226-244. [47] Queueing Systems, Leonard Kleinrock, New York, Wiley, 1974-76. [48] W. E. Leland, et al., \On the self-similar nature of ethernet trac," IEEE/ACM Trans. on Networking, Vol. 2, No. 1, February 1994, p. 1-15.

BIBLIOGRAPHY

178

[49] B. B. Mandelbrot, \Self-Similar Error Clusters in Communication Systems and the Concept of Conditional Stationarity," IEEE Trans. on Communications Technology, Vol. 13 pp. 71-90, 1965. [50] Simulation Modeling and Analysis, Averill M. Law and W. David Kelton, McGraw Hill, New York, 1991. [51] T. Brown, S. Sazzad, C. Schroeder, P. Cantrell and J. Gibson, \Packet Video for Heterogeneous Networks using CU-SeeMe," Proceedings of the IEEE International Conference on Image Processing, Lausanne, Switzerland, 1996, pp. 9-12. [52] W. R. Stevens, TCP/IP Illustrated Volume 1, Addison Wesley, Menlo Park, CA. [53] M. Barton and T. R. Hsing, \Architecture for Wireless ATM Networks," PIMRC' 95, September 27-29, 1995, Toronto, Canada. [54] C. Huang et al, \Fast Simulation for Self-Similar Trac in ATM Networks," Proceedings of IEEE ICC 95, June 1995, Seattle, WA., USA. [55] C. Huang et al, \Modeling and Simulation of Self-Similar Variable Bit Rate Compressed Video: A Uni ed Approach," ACM SIGCOMM 95, August 1995, Cambridge, MA., USA. [56] C. Huang et al, \DTMW: A New Congestion Control Scheme for Long Range Dependent Trac," Proceedings of ITC 15, June 1997, Washington, DC., USA. [57] M. Prycker, Asynchronous Transfer Mode: Solution for Broadband ISDN, New York: E. Horwood, 1993. [58] A. S. Acampora and M. Naghshineh, \An architecture and methodology for mobile-executed cell hand-o in wireless ATM networks," Proceedings of International Zurich Seminar on Digital Communications, Zurich, Switzerland, 8-11 March 1994.

BIBLIOGRAPHY

179

[59] K. Keeton et al., \Providing Connection-oriented Network Services to Mobile Hosts," USENIX Symposium on Mobile & Location-Independent Computing , Aug. 2-3, 1993, Cambridge, MA., USA. [60] Ian M. Leslie, et al., \ATM Everywhere," IEEE Network, pp. 40-46, March 1993. [61] Arthur Miller, \From here to ATM," IEEE Spectrum, pp. 20-24, June 1994. [62] Bora A. Akyol and Donald C. Cox, \Handling Mobility in a Wireless ATM Network," INFOCOM' 96, March 24-28, 1996, San Francisco, CA., USA. [63] Bora A. Akyol and Donald C. Cox, \Rerouting for Hando in a Wireless ATM Network," ICUPC' 96, September 28-October 2, 1996, Cambridge, MA., USA. [64] Bora A. Akyol and Donald C. Cox, \Signaling Alternatives in a Wireless ATM Network," IEEE J. on Selected Areas in Communications, Vol. 15, No. 1, January 1997, pp. 35-49. [65] Bora A. Akyol, \Modeling of A Personal Wireless Communication Network Signaling System," Internal Report, Stanford University, November 8, 1994, available by request from the author at [email protected]. [66] Bora A. Akyol, \The Overlay Signaling Approach with Permanent Virtual Circuits," Internal Report, October 25, 1996, available by request from the author at [email protected]. [67] B. A. Akyol and C-K Toh, \A Survey of Handover Techniques for Wireless ATM Networks," submitted to Journal of Wireless Information Networks.

AN ARCHITECTURE FOR A FUTURE WIRELESS ATM NETWORK

AN ARCHITECTURE FOR A FUTURE WIRELESS ATM NETWORK

Suggest Documents

an architecture for a future wireless atm network - CiteSeerX

SWAN: An Indoor Wireless ATM Network - CiteSeerX

SWAN: An Indoor Wireless ATM Network - CiteSeerX

An architecture of ATM-based PCS network

the wireless atm architecture - Radioengineering

An Efficient Location Management for Hybrid Wireless ATM Network ...

A Hierarchical Wireless Network Architecture for ...

HOMEPLANE: An Architecture for a Wireless Home Area Network with ...

A signaling architecture for wireless ATM access ... - Springer Link

A signaling architecture for wireless ATM access ... - Springer Link

Rerouting for Hando in a Wireless ATM Network - CiteSeerX

A Control and Management Network for Wireless ATM Systems

Data Link Control Protocols for a Wireless ATM Network

A Control and Management Network for Wireless ATM Systems - ITTC

An architecture of ATM-based PCS network: Signaling ... - CiteSeerX

Mobility Management in an IP-based Wireless ATM Network

Satellite ATM network architecture - Department of Computer ...

Wireless ATM

An SDN-based Network Architecture for Extremely Dense Wireless ...

An RTOS-based Architecture for Industrial Wireless Sensor Network ...

Wireless Network Architecture for Long range

Wireless Network Architecture for Long range

Wireless Sensor Network Architecture - ipcsit

An IRAM-Based Architecture for a Single- Chip ATM Switch