Appeared as part of "Mobility Management in Next Generation Wireless Systems," Proceedings of the IEEE, Vol. 87, No. 8, pp. 1347{84, August 1999. (pages 1365{76)
Mobility Management Protocols for Wireless ATM Networks Janise Y. McNair School of Electrical and Computer Engineering Georgia Institute of Technology Atlanta, GA 30332 Email:
[email protected] Tel: (404)-894-6616
Abstract This paper presents a survey of mobility management protocols for Wireless Asynchronous Transfer Mode (WATM) networks. Basic architectural assumptions for WATM networks are presented, followed by an overview of location registration, call delivery, and terminal paging techniques. Next, a collection of protocols for hando connection re-routing is examined. The paper concludes with a discussion of standardization activity within the ATM Forum. Key Words: ATM Forum, Hando , Location Management, Mobility Management, Wireless ATM
1
1 Introduction Wireless networks have evolved beyond the well-known and widely used voice and paging services. The commercial success of cellular telephony has created a great demand for mobile communications and computing, along with an even greater nancial return [23]. On the other hand, wireline telephony networks have evolved to integrated services networks that can support a variety of trac types through reliable transport and service guarantees. One such network is the Asynchronous Transfer Mode (ATM) network. ATM networks transport high volumes of voice, data, video, and video conferencing trac according to Quality-of-Service (QoS) guarantees. The result of the concurrent successes of these two evolving telecommunication systems has been a research e ort to develop a wireless ATM network (WATM) that can provide a user with tetherless, multimedia communication and computing services, supported by QoS guarantees. One of the most important areas of this research e ort is the development of protocols for mobility management [30]. Mobility management enables a wireless network to locate roaming terminals for call delivery and then maintain the connections as the terminals change their access points to the network. It allows users to roam while simultaneously o ering them incoming calls and support for calls-in-progress. This mechanism can be segmented into two mobility governing processes: location management and hando management [21].
Location Management Location management is a two-stage process that enables the network
to locate the mobile user for call delivery. The rst stage is called location registration (or location update). In this stage, the mobile user periodically noti es the network of its current location and allows the network to update its user location pro le. The second stage is called call delivery. This stage involves querying the network for the user location pro le in order to route incoming calls to the mobile user's current location. Location management a ects issues such as security and paging. The location management operations and their respective research issues are shown in Figure 1. RESEARCH ISSUES LOCATION MANAGEMENT
AUTHENTICATION SIGNALING TRAFFIC CENTRALIZED VS. DISTRIBUTED DATABASE ARCHITECTURES
LOCATION REGISTRATION
CALL DELIVERY TERMINAL PAGING
(LOCATION UPDATE) DELAY CONSTRAINTS
Figure 1: Location Management Operations 2
Current methods for location management use a centralized database architecture and signaling protocols such as terminal paging to track mobile users [6]. As the number of mobile subscribers increases from 200 million worldwide to a projected 500-600 million [1], wireless resources must be able to accommodate greater and greater trac requirements. Thus, it is important to develop protocols that e ectively support a continuously increasing subscriber population. In this paper, we discuss the algorithms that address such issues through new database and signaling structures that aid in the ecient transfer of location information.
Hando Management Hando (or handover) management enables the network to maintain a
user's connection as the mobile host continues to move and change its access point to the network. The three-stage process for hando begins with initiation, where either the user, a network agent, or changing network conditions identify the need for hando . The second stage is new connection generation. Under Network-Controlled Hando (NCHO), or Mobile-Assisted Hando (MAHO), the network generates a new connection, nding new resources for the hando and performing any additional routing operations. For Mobile-Controlled Hando (MCHO), the MT nds the new resources and the network approves. The nal stage is data ow control, where the delivery of the data from the old connection path to the new connection path is maintained according to agreed-upon Quality-of-Service (QoS) guarantees. The hando management operations and issues are illustrated in Figure 2. RESEARCH ISSUES HANDOFF MANAGEMENT
PACKET PROCESSING SIGNALING LOAD INITIATION
NEW CONNECTION GENERATION
DATA FLOW CONTROL
ROUTE OPTIMIZATION
RESOURCE MANAGEMENT USER MOVEMENT
RESOURCE ALLOCATION NETWORK CONDITIONS
BUFFERING/ SEQUENCING CONNECTION ROUTING
EVALUATION OF ALGORITHMS MULTICAST
SERVICE GUARANTEES
Figure 2: Hando Management Operations These operations concern two types of hando : intra-cell and inter-cell. Intra-cell hando occurs when the quality of the radio channel has deteriorated beyond some threshold, or when there is a re-assignment of wireless cell resources. A gradual deterioration is the \with hint" condition, while a sudden failure represents \without hint" hando . Inter-cell hando occurs when 3
the mobile user has moved into a new cell and requires the initial assignment of wireless resources in that service area in order to transfer its connection. Two additional qualities are used to describe the transfer: soft handover and hard handover. Soft handover occurs when the mobile terminal's connection is passed to the target radio port without interrupting connection from the current serving radio port. Thus, the mobile terminal communicates with two radio ports simultaneously, treating the two connections as multipath signals to be combined at the mobile. Hard handover refers to passing the connection between disjointed radio systems, frequency assignments, or air interface characteristics. A hard handover is a \break-before-make" process at the air interface. The discussion of hando management includes issues such as routing, Quality-of-Service (QoS), and resource management. In this paper, we present a survey of mobility management protocols for WATM networks. In Section 2, we present the basic architectural assumptions for wireless ATM networks. In Section 3, we discuss protocols for location registration, call delivery, and terminal paging. Section 4 describes hando re-routing algorithms. Finally, Section 5 concludes the paper with a discussion of ATM Forum activity.
4
2 Wireless Network Architecture The mobile user accesses the WATM network using a device called a Mobile Terminal (MT). In general, the WATM network that is accessed is based on one of two types of wireless network architectures: ad hoc and infra-structured [12]. Ad hoc architectures are concerned with peerto-peer communications, as shown in Figure 3(a). Each MT acts as a router, and one or more MTs performs control functions, such as creation/deletion of wireless connections between peers, resource management, and QoS management. To access the outside world, i.e., the ATM backbone network, additional internetworking support at the MTs is required. MTs with Controllers
ATM BACKBONE NETWORK ATM Switch
ATM Switch
Wireless Connections to Base Stations
Peer-to-Peer Wireline Connections to ATM Switch
Wireless Connections
BS MTs Cell
MT (a) Ad Hoc WATM Architecture
(b) Infra-structure WATM Architecture
Figure 3: Wireless ATM Network Architectures Infra-structured WATM networks, on the other hand, consist of radio coverage areas called cells, as shown in Figure 3(b). Each MT communicates with the Base Station (BS) in its serving cell to gain radio access to other MTs or to the backbone network. The BS converts network signaling trac and user trac to the radio interface for communication with the MTs in its coverage area. Each BS also transmits a broadcast channel carrying the logical cell identi cation which identi es its service area. The MT tunes to the serving broadcast channel associated with the serving BS/cell to receive data concerning the cell organization, paging messages, and a reference signal to be used during cell re-selection and hando . To make or receive calls, the MT tunes to one of the channels assigned by the serving BS. The next level of the infra-structured network is the ATM switching element. The BSs are governed by switching functions which can be implemented in the BSs themselves or in an ATM switch. The ATM switch can be either a traditional wireline switch, or a Mobility-Enhanced ATM (ME-ATM) switch. In either case, the switching element provides the access to the resources of the ATM backbone and manages radio resources. The switches also control communications to 5
any external mobility-governing devices. In the special case of ME-ATM switches, the switch itself governs mobility management procedures. Ad hoc architectures are suited to localized systems, supplying wireless ATM services for conference room or home multimedia applications. As there is little requirement for intercell hando or location management in such systems, ad hoc architecture protocols will not be addressed in this survey. Meanwhile, infra-structured networks can support Networks (LANs) or Wide Area Networks (WANs), delivering both peer and network communications to each MT. The protocols presented in this paper will focus on mobility management for such networks. Mobility management can greatly a ect the utilization of resources within an infra-structured network. For example, the use of database servers for location management allows a traditional wireline ATM switch to be used throughout the network, but can result in an increase of signaling trac. On the other hand, implementing signaling changes to accommodate ME-ATM switches may reduce overall trac, but increases the complexity of the system. Next, we examine location management protocols that consider each of these two techniques.
6
3 Location Management Techniques In the mobility management protocols begin considered for WATM, location management is implemented using two techniques: Location Service and Location Advertisement. Location Service refers to the use of databases to maintain records of the current attachment point of MTs within the network. Once the location information has been obtained, terminal paging is employed to deliver calls to the MT within the service area of its attachment point. Thus, Location Service requires resources for signaling, querying, and paging. Location Advertisement avoids the use of databases by passing location information throughout the network on signaling and/or broadcast messages. As shown in Figure 4, three protocols are examined for Location Service|including a velocity paging scheme, followed by three protocols for Location Advertisement. LOCATION MANAGEMENT
LOCATION SERVICE
TERMINAL PAGING
TWO-TIER DATABASES
LOCATION ADVERTISEMENT
INTEGRATED LOCATION RESOLUTION
VIRTUAL CONNECTION TREE LOCATION REGISTERS
MOBILE PNNI
Figure 4: ATM Location Management Techniques
3.1 Location Service Location Service relies on the use of database servers to store a record of the current position of the mobile. These servers require querying operations, as well as signaling protocols for storage and retrieval [15]. The rst Location Service protocol, Two-Tier Database utilizes a bi-level database structure, while the second technique, Location Registers, uses a hierarchy of databases.
3.1.1 Two-Tier Database The architecture for the two-tier database scheme described in [7] uses bi-level databases that are distributed to zones throughout the network. The zones are each maintained by a zone manager, which controls the zone's location update procedures. The home tier of the zone database stores location information regarding MTs that are permanently registered within that zone, while the second tier of the zone database stores location information on MTs that are visiting the zone. 7
Each MT has a home zone, i.e., a zone in which it is permanently registered. Location management operations are performed according to the present zone location of the MT. Upon entering a new zone, the MT detects the new zone identity broadcast from the BSs. The steps for registration, demonstrated in Figure 5, are as follows: HOME ZONE Home Tier Visitor Tier
(3)
Zone Manager (4)
PREVIOUS ZONE
Zone Manager
(5) Home Tier
CURRENT ZONE
Visitor Tier
(2)
Home Tier Visitor Tier
(1)
Zone Manager
BS MT
Figure 5: Two-Tier Database Scheme 1. The MT transmits a registration request message to the new zone that contains its user identi cation number (UID), authentication data, and the identity of the previous zone. 2. The current zone manager determines the home zone of the MT from the previous zone. 3. The current and home zone managers authenticate the user and update the home user pro le with the new location information. 4. The home zone sends a copy of the pro le to the current zone, which stores the pro le in the Visitor tier of its database. 5. The current zone manager sends a purge message to the previous zone so that the user's pro le is deleted from the previous zone's Visitor tier. Call delivery is achieved by routing the call to the last known zone rst. If the MT has moved since the last update and has been purged from the last known zone, the call is immediately forwarded to the home zone. The home zone's Home Tier database is queried for the current location of the MT, which is forwarded back to the calling switch. The calling switch can then setup a connection to the current serving switch of the MT. 8
The advantage of the two-tier scheme is that it keeps the number of queries low|requiring at most two database lookups for each incoming call to nd the MT. However, the use of a centralized Home Tier database may greatly increase signaling trac and unnecessary connection setup delays if the MT makes several localized moves for an extended period of time. The next protocol attempts to reduce the e ects of long-distance signaling by localizing the transfer of location information within the levels of a database hierarchy.
3.1.2 Location Registers Hierarchy The Location Registers scheme in [28] distributes location databases throughout a hierarchical PNNI (Private Network-to-Network Interface) architecture, as shown in Figure 6. The PNNI procedure is based on a hierarchy of peer groups, each consisting of collections of ATM switches. Each switch can connect to other switches within its peer group. Special switches, designated peer group leaders, can also connect to a higher-ranking leader in the \parent" peer group. Each peer group also has its own database, or Location Register (LR), used to store location information on each of the MTs being serviced by the peer group. The PNNI organization allows the network to route connections to the MT without requiring the parent nodes to have exact location information. Only the lowest referenced peer must record the exact location, and the number of LR updates then corresponds to the level of mobility of the MT. For example, a connection being setup to a MT located at switch A:2:2 in Figure 6 is rst routed according to the highest boundary peer group from switch A. Switch A can then route the connection to its \child" peer group, level A:x, to switch A:2. Finally, the connection is routed by A:2 to the lowest peer group level to switch A:2:2, which resolves the connection to the MT. For movement within the A:2:x peer group, location updates can be localized to only the LR of that peer group. However, a movement from peer group B:1:x to peer group A:2:x requires location registration of a larger scope, and the maintenance of a home LR to store a pointer to the current parent peer position of the MT. To limit signaling for the larger scale moves to the minimum necessary level, the authors implement two scope-limiting parameters, S and L. The S parameter indicates a higher-level peer group boundary for LR queries, while the L parameter designates the lowest group. In Figure 6, the S-level is level 1 (B:1), while the L-level is level 2 (B:2:x). When the MT performs a location update by sending a registration noti cation message to the new BS, this message is relayed to the serving switch, which then stores the MT's location information in the peer group's LR. When the MT powers on or o , this message is relayed up the hierarchy until it reaches the pre-set boundary, S. The S-level register records the entry and then relays the message to the MT's home LR. For movement from position B:1:2 to position 9
Peer Group A
(5) Peer Group Leaders
(MT’s Home LR) A.1
B
Location Register
LEVEL 0
ATM Switches
(4) (former S-level) A.2
B.1
B.2
LEVEL 1
(3)
A.2.1
A.1.1
B.1.1
B.2.1
A.1.3 LEVEL 2
(2) A.2.2 A.1.2
B.1.2
A.1.4
B.2.2
(1)
BS MT
(MT’s Former Position)
Figure 6: WATM Location Registers Scheme
A:2:2, the registration procedure, shown in Figure 6, is as follows: 1. The MT sends a registration noti cation message to the new BS/switch 2. The new switch stores the MT's location information at in the peer group's Location Register 3. The peer group then relays the new location information to the higher level Location Registers for routing, stopping at the common ancestor of the former and current peer groups. 4. In this example, the former S-level is not a common ancestor, so a new S-level is designated to be level 0. 5. The MT's home LR (located at group A:x) is noti ed of the new S-level location for the MT After the updates are complete, the new switch sends a purge message to the previous switch so that the former location can be removed from the LRs. Call delivery is less complicated for this method, since the procedure takes advantage of the hierarchical organization. An incoming call request can be routed to the last known peer group or switch via the S-level LR. If the MT has moved, the last known switch propagates a location request, querying the upstream LRs until the mobile endpoint's address is recognized by 10
an LR that has a pointer to the mobile's current position. Then the request is sent to the L-level LR for that peer group, which resolves the query and sends the location information back to the calling switch. Finally, if an incoming call request reaches the former S-level before being recognized, the former S-level LR forwards the location request directly to the home switch. Since the home LR keeps track of S-level changes for its mobile, the home switch can forward the request directly to the current S-level switch, whose LR points to the current peer group position of the MT. In all of the schemes that employ the use of servers or zones to maintain location information, a method is required to nd the MT once the current zone position has been determined. Terminal paging allows the serving BS or zone manager to deliver calls to a MT that has moved into its service area.
3.1.3 Terminal Paging Research Terminal paging research has largely focused on the general Personal Communication Services Networks and many of these protocols have been outlined for PCS in [6]. The scheme below is a recent sample.
Velocity Paging Scheme The velocity paging scheme outlined in [32], attempts to categorize
the travel of the MT into a velocity class. The class is then used to generate a paging zone|a list of cells to be paged. The velocity classes described for the scheme characterize a range of velocities that has been previously demonstrated by the MT. This quantity can be obtained in two ways. The rst way employs a distance based registration, which is also seen for PCS in [6], wherein the MT registers whenever its distance from the last registered cell has passed a threshold value. The velocity class can then be formed by using the distance threshold value divided by the time between consecutive registration actions. Then an average speed for the MT has been determined. The other way to nd the velocity class of a mobile is to take advantage of the movement-based registration procedure. The movement-based procedure counts the number of times the MT has traversed through a cell. Then, once that number has reached a threshold, the MT must register. The number of movements, divided by the time passed between registrations and multiplied by a Velocity Time Unit would yield the quantity that we desire. The procedure for paging a mobile terminal is outlined next. When the system wants to deliver a call to a standby MT, the system must query a location server. The server supplies the MT's movement pro le, based on the two choices above. It supplies the velocity class index for the MT, the MT's last known location, and the last registration time. The system then uses this information to calculate the maximum distance that the MT could have traveled within the given constraints. Finally, the candidate cells that are within that maximum 11
distance are the rst group to be paged. Alternatives to this scheme, described in [32], include recording the direction of the MT, paging in that direction initially, and then branching out. Additional alternatives to paging are presented for PCS in [16] and [5]. Although the location server methods provide the advantages of simplicity, decreased computation costs and exibility, the method can still require a substantial signaling and database querying load. In addition, terminal paging on the air interface reduces the availability of much-needed radio resources. The next group of protocols take advantage of existing signaling trac by attaching location information to messages that are already being transported as part of the signaling procedures.
3.2 Location Advertisement Location Advertisement refers to the noti cation of appropriate network nodes of the current location of the MT without the need for database processing. The rst method, Mobile PNNI, uses the PNNI architecture described in Section 3.1 by removing the Location Registers and taking advantage of an internal broadcast mechanism [28]. The second method, Destination-Rooted Virtual Connection Trees, advertises location information via provisioned virtual paths [30]. The third and nal method, Integrated Location Resolution extends the signaling framework of ATM with location information elements that incorporate location resolution into the connection setup process [4].
3.2.1 Mobile PNNI This scheme, described in [28], uses the status noti cation procedures of the PNNI network to achieve automatic registration, tracking and locating. The PNNI protocol calls for the exchange of PNNI Topology State Packets (PTSPs) between ATM switches in the same peer group, between the peer group leader and higher level peer groups, and between the peer group leader and its lower level groups. The PTSPs are generated by the peer group leaders and contain information about the topology of the group and the load on each peer switch. In addition, the PTSPs contain \reachability" information, i.e. address and parent peer group information, for each network endpoint. By exploiting PNNI, some level of location information for each mobile endpoint can be propagated throughout the entire network [13]. The registration procedure occurs without the use of a database, since the location information is contained in these PTSP packets. Instead, a mobile is assigned a home switch which tabulates the route to the mobile's current switch location. The routing table is updated whenever the mobile powers on/o or moves to a new switch. In addition, the home switch is responsible for advertising the mobile's new location to the network, by updating the reachability information 12
in the PTSPs. Peer Group Peer Group Leaders
B
A
ATM Switches
(2) A.1
A.2
B.1
(2) (2)
MT’s Home ATM Switch
B.2
(2)
A.1.1
A.2.1 A.1.3
B.1.1
B.2.1
(2) B.1.2 A.2.2 A.1.2
B.2.2
A.1.4
(1) (2) PTSP packets BS MT
(MT’s Former Position)
Figure 7: Mobile PNNI Registration and update procedures are demonstrated in Figure 7 and are implemented in the following two steps: 1. The mobile sends a registration message to the home switch via the mobile's current access point. 2. The home switch and the current peer group leader propagate the new location information in the PTSPs. The home switch must send a message to the former switch to begin forwarding packets. Once the reachability information has had time to propagate, the former switch is noti ed to stop forwarding packets. Since PNNI reachability only identi es the parent peer group of the endpoints, the extended mobile PNNI algorithm includes a scheme to ood the exact location of the mobile through the network, but only to a certain peer level. This update region is referred to by the authors as a neighborhood in [28]. If an MT moves to a new switch in the same neighborhood as its home switch, the exact location is ooded to the home switch, without additional signaling load. Switches outside of the 13
neighborhood only receive the default, parent peer group information for that endpoint, and can then use aggregate information to route calls according to the method outlined in the Location Registers Scheme in Section 3.1. Until the update propagates through the entire neighborhood, the mobile must send its new location information to the mobile's former switch so that its connections can be forwarded. If a mobile endpoint moves to a new switch that is not in its home neighborhood, the mobile must register its new location with the home switch for storage and update. Again, the previous switch must noti ed, and calls must initially be forwarded. The call delivery phase of Mobile PNNI includes no prior connection setup, since each switch can route the call based on the reachability information it has received. A call from inside the same neighborhood can be immediately routed to the correct switch, provided the latest update has occurred. The same is true for call to a mobile that is in its home neighborhood. For all other possibilities, the home switch must route the call to the current switch. In all cases, if the latest reachability update has not had time to propagate, the last known switch forwards the call to the current switch. The propagation of PTSP packets within speci ed levels does eliminate the need for database processing, but it introduces additional complexity. The next scheme employs a more traditional approach to ooding information through a network|broadcasting.
3.2.2 Destination-Rooted Virtual Connection Tree The network architecture for the Destination-Rooted Virtual Connection Tree Scheme [30] is a collection of portable Base Stations (PBSs) connected via provisioned virtual paths forming a connection tree, as shown in Figure 8. The PBSs are equipped with switching capabilities and limited bu ering capabilities. The trees are based on the mobility indications of the MT. Each PBS maintains a running list of resident MTs in its coverage area. The registration process, also shown in Figure 8, can begin under the power on/o condition, or when the MT moves into a new service area. When the MT powers on or o , the MT sends a message to its local (current) PBS. The PBS simply adds/deletes the MT to/from the service list. However, when the MT moves into a new PBS's region, the new PBS must send a de-registration message to the old PBS on behalf of the MT and then enter the MT's identity information into its current list. Call delivery consists of advertising the mobile terminal's identity via a broadcast message from the PBS of the calling terminal. No paging on the air interface is required. The local PBS responds to the broadcast and initiates connection procedures. If there is no response, the connection is rejected based on the assumption that the MT is not registered. For mobile-to- xed 14
Portable Base Station
Cell Boundary
De-registration Message
MT’s Former Position MT
Registration Message
Figure 8: Destination-Rooted Virtual Connection Tree communication, the PBS performs as a switch with routing tables for the xed endpoint. The simplicity of the broadcast method can quickly absorb needed resources depending upon the number of MTs located at each PBS [30]. The nal protocol for this section, adds to call setup signaling to avoid generating additional trac where possible.
3.2.3 Integrated Location Resolution The Integrated Scheme presented in [4] modi es the signaling operations of the ATM call setup process to include indications of the called terminal's current location. In this scheme, the MT is assigned a home switch which governs all information on the mobile's current position within the network. When the MT moves away from the home switch, the MT must rst register its prescence with the new, foreign switch. Then the foreign switch must notify the MT's home switch of the MT's new foreign address. Thus, the location update operation consists of one update propagated to the MT's home switch. For the call delivery phase of location management, all initial call requests are routed immediately to the MT's home switch with a connection SETUP message. Since this scheme assumes that the home switch of each MT is an ME-ATM switch, the SETUP message can include a ag indicating that an upstream entity is an ME-ATM switch. However, since the calling endpoint does not know if the called terminal is mobile or not, any one of the following conditions can exist:
The called terminal is a static, or xed terminal that is permanently attached to its home switch. 15
The called terminal is mobile, but is currently attached to its home switch.
The called terminal is mobile and it is currently away from is home switch.
At the called terminal's NNI/UNI boundary, the home switch must determine the current condition of the terminal. If the called terminal is xed, the home switch sends a CONNECT message to the originating switch. If the called terminal is a mobile attached to its home switch, the home switch sends a CONNECT message to the originating switch that also identi es the connection as mobile to any interested intermediate switches. This allows any ME-ATM switches in the path to prepare for future possible hando s. (1)
endpoint (1)
(3) (4)
(1)
Mobilliy-Enhanced Switch
(3)
(4)
(2) MT’s Home Switch
MT’s Foreign Switch
BS MT
Figure 9: Integrated Location Resolution The call delivery scheme for a mobile terminal that is currently away from its home switch is shown in Figure 9 and is implemented as follows: 1. The calling endpoint issues a connection SETUP message to the MT's home switch. 2. The home switch determines that the terminal is away from home. 3. The MT's home switch sends a RELEASE message to the originating switch. This message must indicate the MT's current foreign address and also identify the connection as mobile. 4. A switch in the original SETUP path establishes a new path for the connection to the MT and sends a new SETUP message to the MT's foreign address. This new message must include the MT's home address. 16
Location Service
Location Advertisement
Advantages
Disadvantages
Advantages
Disadvantages
exibility scalability
database admin no paging no scalability signaling load no database wasted bandwidth admin for broadcasts
Table 1: Comparison of Location Management Approaches If in any of these cases, the calling endpoint is an MT, the home switch for the calling MT can identify the connection as mobile and include the calling MT's home address during the connection SETUP phase.
3.3 Comparison of Location Management Techniques The location management techniques presented in this paper present a trade o to the implementation of mobility. The Location Service protocols seek to maintain a level of exibility and modularity for obtaining location information, while Location Advertisement seeks to limit the use of signaling resources. The rst Location Service protocol, the Two-Tier Database Scheme, is advantageous in a geographically partitioned network [7]. It has a straight forward approach with a tracking mechanism that requires at most only two data queries to nd the mobile. Thus, the advantages to this system are its simplicity and its limited computation costs. If querying can be reduced to one query at the last known visitor database, the delay in locating the MT can be reduced dramatically. Some disadvantages to the scheme are that it requires each terminal to be recognized as either mobile or xed in order to invoke location resolution. It also increases the overall signaling load, since location updates may require \long distance" signaling to the home tier databases. On the other hand, the hierarchical structure of the Location Registers method provides better organization for the distribution of location information. The tracking and location costs are localized, depending on the scope parameters. By adjusting the range of these parameters, the method can exercise exibility, allowing the user to change coverage areas, service providers, or home locations without a signi cant increase in administrative cost. However, these improvements result in an increase in complexity. Thus, in general, the Location Service protocols o er a level of scalability, and limited computation costs, at the expense of signaling load, paging costs, and database administration. 17
Each of the Location Advertisement schemes eliminate the need for servers and paging by taking advantage of network signaling protocols. The Destination-Rooted Virtual Connection Tree scheme is the least complex, requiring only the broadcast registration and de-registration messages to ood the network. However, the implementation is not scalable and is an inecient use of wired network bandwidth. Mobile PNNI is the most complex, since the PTSP packets must be ooded according to the scope limiting parameters. The Integrated Scheme adds simple signaling items to the connection setup process and leaves some opportunity for optimization. Thus, some of the disadvantages for Location Advertisement are a lack of scalability, additional consumption of network bandwidth for broadcast location queries. Table 1 shows the comparison of location mangement schemes. Mobility-Enhanced ATM switches have a role in the managing the transparency of mobile terminals and also the management of signaling trac for mobile location message routing. These switches also have a role in routing messages and delivered connections for optimality. When WATM networks are able to incorporate all ME-ATM switches, optimal full connection re-routing is possible. This issue not only applies to newly established calls, but also applies to calls-inprogress that have been handed-o to a new access point. In the next section, we discuss full connection re-routing along with several other protocol techniques for hando management for WATM.
18
4 Hando Management Research As mentioned in the introduction, hando (or handover) management controls the process of maintaining each connection at a certain level of QoS as the MT moves into di erent service areas [22]. As illustrated in Figure 10, current proposed protocols for WATM can be grouped into four categories: Full Connection Re-routing, Route Augmentation, Partial Connection Re-routing, and Multicast Connection Re-routing. Full Connection Re-routing maintains the connection by establishing a completely new route for each hando |as if it were a brand new call. Route augmentation simply extends the original connection with a hop to the MT's next location. Partial Connection Re-routing re-establishes certain segments of the original connection, while preserving the remainder. Finally, Multicast Connection Re-routing combines the former three techniques, but includes the maintenance of potential hando connection routes to support the original connection, reducing the time spent in nding a new route for hando [8]. HANDOFF MANAGEMENT
FULL CONNECTION RE-ROUTING
ROUTE AUGMENTATION
InterWorking Devices Connection Re-route
InterWorking Devices Connection Extension
MULTICAST CONNECTION RE-ROUTING
PARTIAL CONNECTION RE-ROUTING
Nearest Common Node Re-routing
Virtual Connection Tree Re-routing
Hybrid Connection Re-routing
Homing Base Station Re-routing
Figure 10: WATM Hando Management Techniques
4.1 Full Connection Re-Routing Full connection re-establishment is the most optimal and the simplest re-routing technique in that all of the VCs in the connection path from the source to the previous switch are cleared and new VCs are established from the source to the new switch [25]. It can be implemented by treating the hando connection as a newly admitted call, or by employing network elements that perform 19
the mobility functions for that connection independent of the switches.
4.1.1 InterWorking Devices { Full Connection Re-route The Connection Re-route Algorithm for InterWorking Devices outlined in [25] manages hando using external processors that isolate the mobility management from the xed network. An overlay of these processors, called InterWorking Devices (IWDs), is placed throughout the xed network| one IWD corresponding to each ATM switch, as shown in Figure 11(a). The routing and signaling procedures outlined in [25] for Full Connection Re-route are illustrated in Figures 11(a) and (b), and listed below: 1. After registration, the MT informs the new target switch of the identity of the original connection, including a unique identi er for the connection and the address of the former IWD. 2. The target switch then sends the hando request message to its IWD. 3. (3i) The target IWD then forwards this request to the calling IWD. This message is sent via the IWD overlay and the connection ID. 4. (4i) The calling IWD and target IWD set up a new connection to the MT. 5. (5i) To clear the original path, the calling IWD sends a clear message along the old path toward the former switch. The Full Connection technique is always optimal, but latent, as the new route must be calculated for each hando . Route Augmentation, on the other hand, provides a level of speed and simplicity that can potentially reduce hando latency and signaling costs.
4.2 Route Augmentation The route augmentation technique is the simplest method to achieve hando since it requires no cell sequencing, little bu ering, and no optimization. It consists of augmenting the existing connection route with an added route from the last position to the current position of the MT.
4.2.1 InterWorkingDevices { Connection Extension In this scheme outlined in [25], the hando operation uses the IWDs as anchors through which a hando connection can be extended to the next hop. The IWD architecture and the route augmentation signal ow are depicted in Figure 11(a) and (c) as follows: 20
Calling IWD
Calling Switch
(5i)
(3i) (4i)
BS MT
CALLING NODE
Target IWD
Target Switch
Former
(3j) (4j)
IWD
Former Switch
(2)
(5j)
(1) TARGET NODE
BS
BS MT
FORMER NODE
(a)
Target Switch
MT
ID open Conn (1)
Target IWD
Request Handoff (2)
Calling IWD
Request Handoff (3i)
Conn Setup
Conn Setup
Former Switch
MT
Target Switch
ID open Conn (1)
(5i) Clear Old VCs
Target IWD
Former IWD
Re-direct Request (3j)
Conn Setup (4i)
Conn Setup Clear Old VCs
Former Switch
(2) Request Handoff
Conn Setup
(b)
Clear Old VCs (5j)
VC Setup (4j) Clear Old VCs
(c)
Figure 11: InterWorking Devices Hando Scheme (a) Architecture (b) Connection Re-route Signal Flow (c) Connection Extension Signal Flow 1. Again the MT informs the target switch of the identity of the original connection, including a unique identi er for the connection and the address of the former IWD. This identi es the former IWD as the anchor. 2. The target switch then forwards the hando request message to its IWD. 3. (3j) The target IWD sends a re-direct request to the former IWD. 4. (4j) The former IWD responds by bridging the VC connection to the target IWD. The target IWD can notify the target switch to set up the new connection to the MT. 5. (5j) Finally, the former switch clears its connections to the MT. The full connection technique establishes an optimal route at the expense of signaling and resource management, while route augmentation simpli es the process at the expense of eciency. Thus, current research activity focuses on partial connection re-routing to preserve 21
limited optimization and simplicity and yet reduce the negative e ects of resource waste and ineciency.
4.3 Partial Connection Re-routing The partial connection re-routing technique attempts to route connections by preserving some portions of the original route for simplicity and re-routing other portions for optimality.
4.3.1 Nearest Common Node Algorithm The Nearest Common Node Re-routing (NCNR) algorithm presented in [8] routes connections according to the residing zone of the MT. Hando within a zone constitutes a VC table update. For hando between zones, the algorithm attempts to bridge the connection at the nearest WATM network node that is common to both of the zones involved in the hando transaction. In the tree topology, common refers to two nodes branching from the same point. In a hierarchy, the common point of two zones is a higher node which uses separate paths to access each zone. Calling Switch
Fixed Endpoint
(1)
(3) NCN (Nearest Common Node)
(2)
(4) (1) (2) (1) (4)
Former Switch
Target Switch
ZONE A
ZONE B
BS MT
Figure 12: Nearest Common Node Re-routing Procedure Hando from zone A to zone B begins with the zone manager of zone A checking for a 22
direct physical link between the two zones. If either zone is a "parent" of the other, the parent zone acts as an anchor for the hando procedure. In this case, if A is the parent zone, A becomes an ATM switch in the connection path, while if B is the parent zone, B can delete A from the connection path altogether. On the other hand, if A and B are not directly linked, a hando from zone A to zone B is conducted as illustrated in Figure 12 and listed below: 1. Zone manager A transmits a hando start message to the calling switch. This message includes the ATM addresses of the MT, the target switch, and the former switch. 2. Each switch in the original connection path uses the three ATM addresses in the hando start message to determine if it is the nearest common node (NCN). (The NCN has 3 di erent egress ports for the 3 addresses.) 3. Since the above step repeats itself until the NCN is determined, the switch that recognizes itself as the NCN stops this process by setting the NCN bit in the hando start message. 4. The NCN initiates a connection to the target zone by forwarding a re-route message to the switches in that path. The target zone receives the re-route message and replies with an ACK to the former zone, thereby completing the re-route process. Until the radio-level hando is complete, the NCN forwards the connection data to zones A and B. Afterward, the NCN clears its connection to zone A.
4.3.2 Hybrid Connection Algorithm In the Hybrid Connection Algorithm [27], the target BS must determine whether the hando is intra-cluster or inter-cluster. A cluster is a collection of BSs connected to a common cluster switch. If the hando is intra-cluster, the cluster switch can perform the hando as the cross over switch (COS). Otherwise, a COS discovery process must be initiated. This procedure is shown in Figure 13 and outlined below: 1. The MT sends a hando hint message to the former switch. 2. The former switch sends a hando invoke message (containing the MT's connection list) to the target switch. 3. A process is invoked to determine the cross-over switch [27]. 4. Partial paths are setup from the COS to the target switch. 23
Calling Switch
Fixed Endpoint
(3) COS (Cross Over Switch)
(2)
(2)
(4)
(2)
(2)
(4)
Former Switch
Target Cluster
Former Cluster
Target Switch
(1)
BS MT
Figure 13: Hybrid Connection Re-routing Procedure By the time the MT leaves the overlap region and fully enters the new region, the new path has been established. The MT can then send a greet message to the target BS and the target switch can send a redirect to the COS in order become a part of the new connection path. Finally, the COS informs the former switch to disconnect the old partial connection path. For hando due to link failure, some cells will be lost until the MT detects the failure. Bu ering at the MT and at the COS is used to re-gain the appropriate hando connection. Partial connection re-routing provides better resource utilization while reducing signaling, but it requires algorithms for bu ering and cell sequencing. It also requires the computation of the Nearest Common Node or the Cross Over Switch. The method described next incorporates the use of multiple hando paths in order to minimize cell loss and bu ering needs.
4.4 Multicast Connection Re-routing Multicast re-establishment combines the ideas discussed previously in a hybrid fashion, but also introduces the idea of maintaining the potential hando connections in addition to the original connection. Then, since several new routes are available, there is little network time spent in selecting a new hando route [11].
24
ATM BACKBONE
Virtual Connection Tree Root (3) Routing Table Update
(2)
(2)
Former Switch
Target Switch
(2) (1)
BS MT
Figure 14: Virtual Connection Tree Algorithm
4.4.1 Virtual Connection Tree Algorithm The virtual connection tree algorithm in [2] is based on a hierarchical collection of ATM switching nodes attached to the xed network, with links extending to BSs or Access Points (APs). The root of the tree is a xed switching node connected to the backbone network, while the leaves are the APs. Each mobile connection is assigned a set of Virtual Connection Numbers (VCNs) that are used to identify a set of paths from the root to one leaf. Only one path is operational at a time, to route a connection from the root to the MT. The authors of this protocol outlined hando procedures for two cases: hando within the same tree and hando between trees. Figure 14 demonstrates hando within the same tree: 1. The MT transmits cells with a new VCN, corresponding to the previously set-aside path for the target BS 2. The target BS then activates the VCN path by using it to transmit packets from the MT to the root of the tree. 3. When the cells arrive at the root with a new VCN, the routing table at the root switch is updated with the new BS location of the MT. Incoming cells can then routed to the MT via the new path. 25
From the root of the tree, the connection is routed to the endpoint using the backbone network. To hando between trees, the MT's connection reaches the tree boundary and requires entry into the next virtual connection tree. Then the MT must seek admission to the new connection tree, as if it were a new call to be admitted to the network. By using this method, connections of mobile terminals within a geographical mobile access region or tree may be handed over to any other BS within that area without involving the network processor. The larger the region, the greater the likelihood that the mobile will remain within the area for the duration of a call. SOURCE HOME PBS FOR A
DESTINATION HOME PBS FOR B
MT A
MT B
PBS Cell
Figure 15: Homing Algorithm Hando Routing Procedure
4.4.2 Homing Algorithm The Homing Algorithm presented in [30] uses a collection of portable Base Stations (PBSs) connected in a tree of provisioned virtual paths, as described previously for the destination-rooted virtual connection tree location management algorithm and shown again in Figure 15. A home PBS is designated for each mobile terminal accessing the tree. The home station serves as an anchor for each endpoint of the mobile connection and manages the packet transfer during hando . For MT A transmitting to MT B, packets are rst transmitted to the source home PBS for A. Then packets are routed to the destination home PBS for B via pre-established VPs. The home PBS for B delivers the packets directly to MT B. MT A can then execute a hando by transmitting packets to the target PBS. The packets will be routed from the target PBS to the Source Home PBS for A, where the rest of the connection to MT B continues unchanged. The source home-to-destination home connection remains relatively stable due to the anchoring scheme. Thus, if both MT A and MT B have each moved to new PBSs, packets are rst sent to the home station, which is able to forward the packets to the new PBS location. Finally, to 26
improve network utilization, the home PBSs are periodically updated, according to the movement patterns of the MT. For example, if MT A remained at its new PBS for a prescribed period of time, the network would undergo some process to designate the new PBS as the new home PBS for A.
4.5 Comparison of Hando Management Techniques Connection Re-routing schemes have a fundamental e ect on the ability of the WATM network to maintain its ATM-like functions. They determine the whether the network can implement mobile services transparently or with changes to the existing ATM signaling and switch performance. In addition, These schemes a ect the ability to support QoS for connection issues such as cell loss, cell sequencing, delays|due to failures, hando latency, or re-transmission, and jitter. We present here a summary of the hando protocol performance. Detailed analysis can be found in [9]. The rst two methods, Full Connection Re-routing and Connection Extension are at two extremes, while Partial and Multicast Connection Re-routing attempt to achieve an optimal compromise. The Full Connection Re-routing with InterWorking Devices method allows for the transparent support of mobility in that it does not require changing the existing signaling mechanisms. It ensures the delivery of services and inherently performs QoS re-negotiation as it determines an entirely new optimal route with appropriate resources. However, The process introduces unnecessary delays in the hando operation and can lose its level of transparency when the end host must be noti ed for re-calculation of the entire route. Connection Extension is a faster method that preserves the transparency of the InterWorking Devices and ensures cell sequence preservation, but the routing will not be optimized after each hando and therefore can result in loops or other inecient connection patterns. Another technique that does not make ecient use of connections is Multicast Connection Re-routing. These schemes maintain the speed advantage of Connection Extension, but introduce higher bu ering and network bandwidth preallocation needs. These unnecessary connections are eliminated by the Partial Connection Re-routing Technique. It preserves cell sequencing and minimizes the resources required for re-routing. In addition, the Nearest Common Node Rerouting protocol addresses the routing needs of various trac types. However, the cost of these improvements are an increase in processing at the switch and the need for all switches to become mobility enhanced. Table 2 lists the advantages and disadvantages for the Connection Re-routing Techniques. Future research for WATM involves simulating and evaluating connection routing schemes in order to develop signaling protocols that facilitate mobility management and to quantify their e ect on call dropping rates, call blocking rates, cell sequencing, connection setup latency and hando latency [20, 33]. An algorithm for optimizing the route of a connection that becomes sub27
Full
Extension Partial Advantages
optimal route
fast
existing methodology
maintains cell sequence
maintains cell sequence reduced resource utilization
Multicast fast maintains cell sequence
Disadvantages
slow
wastes bandwidth
complex
inecient resource re-assignment
inecient connection route
added switch processing reqs
added bu ering requirements bandwidth pre-allocation
Table 2: Comparison of Hando Management Approaches optimal is presented in [13]. The ATM Forum has been reviewing proposed protocols for mobility management that will become a standard for the modular development of WATM networks [24]. In the next section, we discuss the protocols under consideration for standardization by the ATM Forum, along with remaining open problems.
28
5 ATM Forum Activity The WATM Working Group within the ATM Forum continues to evaluate and re ne existing protocols to establish a standard for WATM networking. The group now focuses on combining techniques from di erent schemes and allowing some exibility of method in order to eciently integrate WATM with existing wireline ATM network infrastructures [24].
Location Management The two protocols under current consideration to be standardized by
the ATM Forum are the Extended Location Registers Scheme and the Extended Mobile PNNI Scheme [14]. Addressing is a concern for standardizing the location registers approach. A strategy is needed to assign permanent addresses to mobile terminals. One suggestion is to split the ATM Network Service Access Point (NSAP) address space into mobile NSAPs and xed terminal NSAPs in order to allow a calling party's switch to generate a location request before initiating connection setup [29]. Another is to create a set of addresses that identi es the level of mobility. Small-scale mobility would represent a limited area de ned by the peer group of the enpoint's current node. Large-scale would represent mobility between peer groups. Continuing issues include security in tracking and locating the mobile terminal, whether to have a continuous active VC from the current node to the home node, and the eciency of continuous ush and re-advertisement [10].
Hando Management The ATM Forum has also been working to combine the main concepts
of the proposed hando schemes in order to nd a feasible yet ecient method of re-routing for hando [24]. Both path re-routing and path-extension require two phases|fast extension and then route optimization. One extreme is the establishment of an entirely new connection|providing an optimal route, but at the expense of latency. The extreme at the other end is connection extension, which is simply a static partial connection re-route policy, where the previous BS is always the designated NCN or COS [31]. This method provides the fastest extension, but is a wasteful use of network resources. A hando signaling framework is needed that can accommodate a number of connection re-routing mechanisms [24]. One consideration is a multicast technique where the mobile would connect to two or more access points at a time, in order to help maintain seamlessness for handover and to possibly reduce bu ering needs [26]. Other challenging issues that remain for the ATM Forum are [24]:
QoS: De ne service classes for mobile connections, devise mechanisms to adapt to QoS
Re-Routing Connections: Develop algorithms for nding new route options, create sig-
degradation, and develop new signaling features, such as QoS re-negotiation.
naling protocols for recon guring the connection path and determine the feasibility of proposed solutions. 29
Point-to-Multipoint: Develop protocols that address re-routing a mobile terminal's point-
Mobile-to-Mobile Hando : Further address changes to existing protocols in order to
Optimization: Develop ecient methods that allow an existing mobile connection to be
to-multipoint connections.
support connection routing and QoS for a mobile-to-mobile connection. periodically re-routed for an optimal connection path.
30
6 Conclusion Mobility management for WATM deals with transitioning from ATM cell transport based upon widely available resources over wireline networks to cell transport based upon the limited and relatively unreliable resources over the wireless channel. Thus, it requires the investigation of location and hando management protocols that a ect ecient use of these resources. The integration of the wireless ATM network into current systems also requires that the introduced protocols have a limited complexity that allows these protocols to be implemented modularly. The nal important step lies in developing evaluation methods for the current Location Service, Location Advertisement, Terminal Paging, and Connection Re-routing schemes in order to quantify the ability of each scheme to be implemented over existing wireline ATM networks, as well as their performance with resource management and QoS issues. The Wireless ATM Working Group of the ATM Forum continues to lead the research e ort with plans to release the WATM Spec 1.0 and the Wireless Interworking speci cation within the next year.
31
References [1] Cellular Telecommunications Industry Association, \The World of Wireless Communications," < http : ==www:wow ? com:com=consumer=faqs= >, Accessed December 16, 1998. [2] A. Acampora, \An architecture and methodology for mobile-executed hando in cellular ATM networks," IEEE Journal on Selected Areas in Communications, Vol. 12, No. 8, pp. 1365{1375, October 1994. [3] A. Acampora, \Wireless ATM: a perspective on issues and prospects," IEEE Personal Communications, pp. 8{17, August 1996. [4] A. Acharya, J. Li, B. Rajagopalan, and D. Raychaudhuri, \Mobility Management in Wireless ATM Networks," IEEE Communications, Vol. 35, No. 11, November 1997. [5] I.F. Akyildiz, J.S.M. Ho, and Y.B. Lin, \Movement-Based Location Update and Selective Paging for PCS Networks," IEEE/ACM Transactions on Networking, Vol. 4, No. 4, pp. 629{636, August 1996. [6] I.F. Akyildiz and J.S.M. Ho, \On Location Management for Personal Communications Networks," IEEE Communications Magazine, Vol. 34, No. 9, pp.135{146, September 1996. [7] B. Akyol and D. Cox, \Handling mobility in a wireless ATM network," Proc. of IEEE INFOCOM '96, Vol. 3, No. 8, pp. 1405{1413, March 1996. [8] B. Akyol and D. Cox, \Re-routing for hando in a wireless ATM network," IEEE Personal Communications, pp. 26{33, October 1996. [9] Bui A. Banh, Gary J. Anido and Eryk Dutkiewicz, \Handover Re-routing Schemes for Connection Oriented Services in Mobile ATM Networks," Proc. of IEEE INFOCOM '98, April 1998. [10] G. Bautz and J. Johnsson, \Proposal for location management in WATM," ATM Forum / 96-1516. [11] S. Biswas and A. Hopper, \A representative based architecture for handling mobility in connection oriented radio networks," Proc. ICUPC '95, pp. 848{852, November 1995. [12] Y. Du, C. Herrmann, K.P. May, S.N. Hulyalkar and D. Evans, \Wireless ATM LAN With and Without Infrastructure," IEEE Communications Magazine, Vol. 36, No. 4, pp. 90{95, April 1998. [13] G. Dommetry, M. Veeraraghavan and M. Singhal, \Route Optimization in Mobile ATM Networks," Proc. of ACM/IEEE MOBICOM '97, pp. 43{54, October 1997. [14] D. Dykeman, I. Iliadis, P. Scotton, L. Frelechoux, and S. Ray, \PNNI Routing Support for Mobile Networks," ATM Forum / 97-0766. [15] L. van Hauwermeiren, L. Vercauteren, A. Saidi, and T. Landegem, \Requirements for Mobility Support in ATM," Proc. of IEEE GLOBECOM '94, pp. 1691{1695, November 1994. [16] J.S.M. Ho and I.F. Akyildiz, \A Mobile User Location Update and Paging Mechanism Under Delay Constraints," ACM-Baltzer Journal of Wireless Networks, Vol. 1, No. 4, pp. 413{425, December 1995. [17] J.S.M. Ho and I.F. Akyildiz, \Local Anchor Scheme for Reducing Signaling Cost in Personal Communication Networks," IEEE/ACM Transactions on Networking, Vol. 4, No. 5, pp. 709{726, October 1996.
32
[18] J.S.M. Ho and I.F. Akyildiz, \Dynamic Hierarchical Database Architecture for Location Management in PCS Networks," IEEE/ACM Transactions on Networking, Vol. 5, No. 5, pp. 646-661, October 1997. [19] J.S.M. Ho and I.F. Akyildiz, \A Dynamic Mobility Tracking Policy for Wireless Personal Communications Networks," Proc. of IEEE GLOBECOM '95, November 1995. [20] C. Kalmanek, P. Mishra, M. Srivastava, and B. Rajagopalan, \Benchmark Methodology for Evaluation of Wireless ATM VC Rerouting Schemes," ATM Forum / 97-0852. [21] A. S. Krishnakumar, \ATM Without Strings: An Overview of Wireless ATM," IEEE Conference on Personal Wireless Communications (ICPWC '96), pp. 216{221, 1996. [22] B. Li, S. Jiang, and D. Tsang, \Subscriber-Assisted Hando Support in Multimedia PCS," Mobile Computing and Communications Review, Vol. 1, No. 3, pp. 29{36, September 1997. [23] C.G. Omidyar and K. Pahlavan, \Guest Editorial: Introduction to Mobile and Wireless ATM," IEEE Communications Magazine, Vol. 35, No. 11, pp. 30{33, November 1997. [24] B. Rajagopalan, \An Overview of ATM Forum's Wireless ATM Standards Activities," ACM Mobile Computing and Communications Review, Vol. 1, No. 3, September 1997. [25] B. Rajagopalan, \Mobility Management in Integrated Wireless-ATM Networks," ACM-Baltzer Journal of Mobile Networks and Applications (MONET), Vol. 1, No. 3, pp. 273{285, 1996. [26] D. Steer and J. Chow, \Requirements for `Soft' Handover," ATM Forum / 97-0696. [27] C.-K. Toh, \A Hybrid Handover Protocol for Local Area Wireless ATM Networks," ACM-Baltzer Journal of Mobile Networks and Applications (MONET), Vol. 1, No. 3, pp. 313{334, 1996. [28] M. Veeraraghavan and G. Dommetry, \Mobile Location Management in ATM Networks," IEEE Journal on Selected Areas in Communications, Vol. 15, No. 18, October 97, pp. 1437{1454. [29] M. Veeraraghavan and G. Dommetry, \Location Management Update," ATM Forum / 96-1701. [30] M. Veeraraghavan, M. Karol, and K. Eng, \Mobility and Connection Management in a Wireless ATM LAN," IEEE Journal on Selected Areas in Communications, Vol. 15, No. 1, pp. 50{68, January 1997. [31] M. Veeraraghavan, M. Karol, and K. Eng, \A Combined Hando Scheme for Mobile ATM Networks," ATM Forum / 96-1700. [32] G. Wan and E. Lin, \A Dynamic Paging Scheme for Wireless Communication Systems," ACM/IEEE MOBICOM '97, pp. 195{203, 1997. [33] R. Yuan, S.K. Biswas, L. J. French, J. Li, and D. Raychaudhuri, \A Signaling and Control Architecture for Mobility Support," ACM-Baltzer Journal of Mobile Networks and Applications (MONET), Vol. 1, No. 3, pp. 287{298, January 1996.
33
