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

Wireless Networks 6 (2000) 145–159

145

A signaling architecture for wireless ATM access networks Nikos H. Loukas a , Nikos I. Passas a , Lazaros Merakos a and Iakovos S. Venieris b b

a Communication Networks Laboratory, Department of Informatics, University of Athens, 15784 Athens, Greece Electrical and Computer Engineering Department, National Technical University of Athens, 15773 Athens, Greece

A multiservice wireless Asynchronous Transfer Mode (ATM) access system is considered from a signaling protocol viewpoint. In an attempt to generalize and extend results and experiences obtained from the specification, design, and implementation of fixed ATM-based access networks, we extend the concept of the broadband V interface (referred to as VB) for application to wireless ATM access networks. The proposed architecture follows the signaling structure of Broadband ISDN (B-ISDN) User–Network Interface (UNI), thus offering the possibility for integration of the wireless ATM access system into fixed B-ISDN. It is shown that the use of the proposed access signaling architecture provides cost effective implementations without degrading the agreed Quality of Service (QoS), and simplifies call/connection and handover control. The evaluation of the proposed access signaling protocol structure yields results that fall within acceptable ATM signaling performance measures. A performance comparison of our approach with an alternative access signaling configuration is also carried out to quantify the relative gains.

1. Introduction In recent years, the developments in the telecommunication industry have been affected by two major trends: broadband and mobile communications. Broadband communication is mainly driven by new multimedia services and applications, requiring more bandwidth and better Quality of Service (QoS) than that offered by existing network solutions. Asynchronous Transfer Mode (ATM) is considered the major technology for the future Broadband Integrated Services Digital Network (B-ISDN) [11]. On the other hand, the parallel emergence of wireless communications networks, based on new digital technologies, is paving the way to provide multiservice capabilities to mobile terminals. These two trends motivate the integration of wireless access systems with the ATM-based fixed broadband network. The wireless ATM technology combines the advantages of wireless operation and freedom of mobility with the service advantages and QoS guarantees of fixed ATM networks. The main challenge of wireless ATM is to harmonize the development of broadband wireless systems with fiber-optic-based infrastructure B-ISDN/ATM and ATM Local Area Networks (LANs), and offer similar advanced multimedia, multiservice features for the support of time sensitive voice communications, LAN data traffic, video and desktop multimedia applications for the wireless user [40]. Such networks aim to provide true ATM wireless access with high-speed transmission (tens of Mbps), flexible bandwidth allocation, QoS selection and guarantees, which cannot be supported by the early proprietary wireless LAN designs, or the wireless LAN standards being developed by IEEE and ETSI [12,20], which are mainly oriented to support data traffic. The implementation of the aforementioned broadband features over the radio interface is a more involved technical challenge than over wired media, owing to the unreliable  J.C. Baltzer AG, Science Publishers

physical medium and user mobility. However, it is important to aim at system architectures that provide qualitatively similar attributes, even if the complete quantitative equivalence with the fixed ATM network may not be feasible [40]. In this context, a number of efforts are in progress to explore this new technology [2–5,14–16,26,30,32,38,41,46]. Most of them comprise wireless research prototypes that are implementing wireless and mobile ATM, though with different approaches and scope. In this paper, we elaborate on the design of wireless ATM access networks from a signaling protocol viewpoint. The proposed signaling protocol structure is based on the same concepts as those used in the fixed broadband access network configurations, currently being considered in standardization bodies [17,18]. Although the signaling architecture considered here aims at local area wireless ATM access systems, it can be readily extended to fulfill the access signaling requirements of public environment wireless ATM access systems. Current trends in designing the access network (AN) part of fixed B-ISDN aim at concentrating the traffic of a number of different User Network Interfaces (UNIs) and routing this traffic to the appropriate Service Node (SN) through a broadband V interface (referred to as VB), as shown in figure 1 [17]. The main objective in AN design is to provide cost effective implementations without degrading the agreed QoS, while achieving high utilization of network resources. This is reflected both in the reduction of the AN physical equipment and in the limitations imposed on the AN functionality, such as the inability to interpret the full ATM layer control information and signaling. The selection of a low-level operating AN forces the establishment of several internal mechanisms that are used to unambiguously identify the connection an ATM packet belongs to, and to convey only those connection parameters that are absolutely necessary for traffic handling.

146

N.H. Loukas et al. / A signaling architecture for wireless ATM access networks

Figure 1. Generic broadband access network configuration.

In this framework, a fast control protocol running over a universal VB interface has been introduced in [48], which serves a number of AN internal functions while preserving the highest possible degree of transparency to the SN. The protocol is based on the Local exchange Access network Interaction Protocol, (LAIP), which was developed to accommodate the SN-to-AN communication requirements, as identified in the early study and design of the dynamic VB5.2 interface, i.e., the interface between the fixed ATM AN and the SN (see figure 1) [25,47]. In the relevant standardization bodies the presence of such a protocol has been firmly decided and has been given the name Broadband Bearer Channel Control Protocol (B-BCCP) [17]. The services of the VB5.2 control protocol enable the dynamic operation of the access network by conveying the necessary connection-related parameters required for dynamic resource allocation, traffic policing, and routing in the AN, as well as information on the status of the AN before a new connection is accepted by the SN. The signaling access architecture for wireless ATM considered here is an extension of the broadband V interface, where an enhanced version of the VB5.2 control protocol is used to enable the dynamic operation of the AN and to serve the AN internal functions. It is assumed that a mobilityenhanced version of the existing B-ISDN UNI Call Control (CC) signaling [7,23,41] is employed to provide the basic call control function and to support the handoverrelated functions. In addition, pure ATM signaling access techniques, based on Metasignaling [21], are adopted for the unique identification and control of signaling channels. These features allow us to minimize the changes required in the signaling infrastructure used in the wired network, and, in this respect, they can guarantee the integration of the wireless ATM access system with fixed B-ISDN. However, when striving for full integration, the mobile-specific requirements imposed by the radio access part need to be taken into account. In today’s wired ATM environment, the user–network interface is a fixed port that remains stationary throughout the connection lifetime. The current B-ISDN UNI protocol stack [7,23] uses a single protocol over fixed point-to-point or point-to-multipoint interfaces. On the other hand, in the wireless ATM access system under study, mobility causes

the users’ access point to the wired network to change constantly, and the mobile terminal connections must be transferred from access point to access point, through a handover process [1]. The support of the handover functionality assumes that the fixed network of the access part has the capability to dynamically set up and release bearer connections during the call. A well-accepted methodology to support these features is the call and bearer separation at the UNI. The use of the extended VB5.2 interface control protocol for wireless ATM access systems serves for the set-up and reconfiguration of fixed bearer connections of the same call, supporting in this way the call and bearer control separation in the AN part. The performance of the proposed AN signaling protocol design is quantitatively studied via simulation. The simulation results obtained capture the effect of the proposed signaling structure on the performance of call and mobility control in terms of call set-up and mean processing delay for handover, and the maximum number of users the system can accommodate. A comparison of our approach with an alternative access signaling configuration, where all the access network elements realize the full B-ISDN UNI signaling protocol stack, is also carried out to quantify the relative gains. The rest of this paper is organized as follows. In section 2, the system architecture is introduced, which defines the signaling interactions between the main functional entities, and specifies the protocol stacks for the various network elements. Section 3 proposes and discusses techniques to guarantee proper signaling operation, while section 4 presents the call/bearer control protocol model, and the enhanced VB control protocol. Section 5 presents the signaling procedures for handover execution, consistently with the proposed signaling architecture. Section 6 presents the model used to evaluate the proposed architecture, and the simulation analysis and results. Finally, section 7 contains the conclusions.

2. Basic architecture and protocol stacks 2.1. Network architecture The network configuration assumes a number of geographically distant local-area (mainly indoor) wireless (and fixed) ATM access systems, interconnected via a core B-ISDN/ATM network, as shown in figure 2. The ATM Mobile Terminal (AMT) is considered as the equipment of the mobile end user, and contains the wireless ATM radio adapted cards interfacing the air interface (e.g., 20 Mbps for [30]). The Access Point (AP) plays the role of a base station and serves the AMTs in its coverage area through a shared radio channel. APs act as gateways for communication between nearby mobile hosts and the backbone ATM/B-ISDN network via the mobile-specific ATM switch. APs do not have switching capability and can be considered as special ATM multiplexers, located at the end of the


147

Figure 2. Wireless ATM network architecture.

wired ATM network, and have wireless link interfaces to the AMTs. The mobile-specific ATM switch is a standard ATM switch equipped with a workstation (Switch Work Station – SWS). The role of the SWS is to provide the additional functionality needed for handling the mobilityspecific functions (e.g., AMT registration, location update, and handover). In addition, the SWS supports the ATM switch in controlling the AMT originated/terminated calls. At the radio interface of the wireless ATM access system, a dynamic Time Division Multiple Access/Time Division Duplex (TDMA/TDD)-based Medium Access Control (MAC) protocol is assumed, for the flexible and efficient ATM traffic transmission over the shared radio channel. A more generic network configuration would consist of different domains in the same geographical area, where each domain is formed by a SWS and its attached APs. As AMTs are allowed to move freely among cells, the problem of handover inside one domain (intra-SWS), or that of roaming between different domains (inter-SWS) arises. However, since here we are concerned with local area (mainly indoor) wireless networks, this paper focuses on intra-SWS handovers (between two radio ports of the same SWS). 2.2. Protocol architecture Figure 3 illustrates the allocation of the four main functional entities used in the proposed protocol architecture in each of the three network elements of the considered access configuration: the Mobile Call Control Entity (MCCE), the Mobility Management Entity (MME), the Bearer Channel Control Entity (BCCE), and the Radio Channel Control Entity (RCCE). Based on this allocation, the following types

Figure 3. Allocation of main functional entities.

of signaling interaction for the communication of peer entities are recognized: • Mobile Call Control Signaling. This includes an enhanced B-ISDN CC signaling protocol (denoted as Q.2931*), based on the ITU recommendation Q.2931 [23], for the set-up, modification, and release of calls between the AMT and the SWS. The enhancements required in the current signaling standards are related to the support of the handover function (e.g., inclusion of handover-specific messages) [41]. • Mobility Management Signaling. This is responsible for the AMT registration/authentication and tracking procedures. • Bearer Channel Control Signaling. This serves for providing the traffic parameters to the AN, and handles the establishment, modification/reconfiguration, and release of fixed ATM connections between the AP and the SWS.

148


Figure 4. Wireless ATM access network protocol stack.

• Radio Channel Control Signaling. This deals with lowlevel signaling related to the radio interface consisting of messages between the AMT and the AP (MAC and physical layer specific messages). The corresponding protocol stacks of the AMT, AP and SWS are given in figure 4. At the user plane, the AMT is presented with a typical ATM protocol stack on top of a radio-specific physical layer and a MAC layer (denoted as MAC/PHY). The AP acts as a simple interworking unit that extracts the encapsulated ATM cells from the MAC frame, and forwards them to the SWS through a proper ATM virtual connection. The MAC functionality realized at the AP is based on a MAC scheduler, which on the basis of the ATM connection characteristics declared at connection set-up and current transmission requests, allocates the radio bandwidth according to the declared QoS requirements and service type of each connection. Such a mechanism provides a degree of transparency to a subset of broadband/ATM services, and achieves efficient sharing of the scarce radio bandwidth among the mobile users [29,36,40–42]. The SWS realizes the typical B-ISDN protocol functionality of the U-plane. The control plane (C-plane) protocol model considers the AMT and SWS acting as termination points for Mobile Call Control (MCC) and Mobility Management Control (MMC). With this architecture, the APs are intentionally kept simple in terms of call and handover control functionality, as only the Bearer and Radio Channel Control (BCC, RCC) protocols are terminated there. Moreover, it is assumed that the basic Mobility Management functions (registration/authentication and location updating) are performed by a location server/database system, which is part of the SWS. We follow the approach of decoupling location management from connection management [51]. This decoupling permits the independent operation of location update schemes. The detailed study of the Mobility Management Control is beyond the scope of this paper.

traffic coming from a number of different UNIs and routed to the appropriate Service Node (i.e., the SWS) through a broadband V (VB) interface, following in this way the same structure as for fixed broadband access networks (figure 1). In this architecture, the mobile user wishes to receive a wide range of multimedia services with acceptable QoS, while the network operator seeks to reduce the overall infrastructure investment. The objective of a cost-effective AN implementation calls for equipment with limited intelligence up to a degree that the overall performance is kept at acceptable levels. A typical feature of such a system is the conscious decision to avoid the interpretation of signaling information, i.e., only a subset of ATM-layer control and management functions is implemented. Internally, the AN should provide sophisticated traffic handling and resource management mechanisms in order to guarantee that it does not become a service performance bottleneck. In the wireless ATM system considered here, the MAC protocol for the radio interface must be able to offer a set of service classes, based on a flexible and efficient MACembedded scheduling scheme [29,36]. To ensure that the radio channel will not be overloaded and guarantee the QoS of existing connections, each new connection has to pass through a radio-specific Connection Admission Control (CAC) process. Its task will be to allocate the available radio bandwidth, according to bandwidth requirements and agreed QoS parameters. The AP entity that performs the radio bandwidth allocation is referred to as Radio Resource Manager (RRM). Additionally, Usage Parameter Control (UPC) is required to preclude malicious or unintentional misbehavior of some connections, which can affect QoS of other conforming connections. The natural position for such a control mechanism in fixed ATM networks is at a crossing point before the shared medium, where all the connections meet, i.e., an ATM switch. On the contrary, in the wireless ATM network considered here the crossing point for the UPC is the AP, placed after the shared radio medium. The true traffic concentration function realized at the AN requires management per Virtual Channel (VC) in the AP. This, together with the requirement for a simple AN, calls for the employment of a fast Bearer Channel Control protocol over the VB interface, i.e., between the SWS and the APs, referred to as the AP-SWS Control Protocol (ASCP). The ASCP operates in real-time on a connection by connection basis and its simple design makes it considerably faster than usual management protocols [48]. The ASCP, which is presented in section 4, is based on the LAIP, initially proposed for a Passive Optical access Network (PON) in [47], and generalized afterwards in [25,48], for any access network.

2.3. ATM-level access network functions

3. Signaling access/signaling channels

With the described network and access signaling protocol structure, the AP can be seen as a concentrator of

In the signaling access architecture described above, a key problem that has to be solved is the provision of a


unique control or signaling channel between AMT and SWS. The major constraint in applying signaling techniques from the wired ATM world [7,23] to wireless environments is the multi-access/broadcast transmission nature of the air interface. The air interface acts as a shared medium compared to the point-to-point interface ATM is primarily designed for. Usually, an end-system is addressed by the triplet: port number, Virtual Path Identifier (VPI) and Virtual Channel Identifier (VCI). Since a port number is nonexistent in a shared medium, and for signaling a fixed signaling VPI/VCI pair (SVPI = 0, SVCI = 5) is defined, a conflict arises at the AP. The reason is that the AP concentrates the up-link traffic and splits the down-link traffic. To ensure proper signaling operation in the wireless AN, signaling sessions should be uniquely identified in the SWS. Signaling cells, passing from several levels of multiplexing between the AMT and the SWS, should be properly demultiplexed so that they can be reassembled and read in the SWS. In the wireless network configuration under study, connections can be multiplexed in two levels: AMT connections that belong to the same AP area are multiplexed by the MAC protocol, while connections from several APs can be multiplexed by a concentrator/multiplexer leading to a switch port. Assuming that different APs are connected to different ports of the switch (as shown in figure 2), no multiplexing of traffic originating from different APs occurs. Hence, the problem is formulated as the ability to distinguish among different signaling sessions of the AMTs using the same AP. This conflict can be solved by allocating a different SVPI/SVCI pair to each connection between a particular AMT and the SWS. An SVPI/SVCI can be assigned dynamically by using the services of the Metasignaling protocol [21,35]. Metasignaling, as described in Q.2120 [21], is part of the layer management entity, and is used for the assignment of a unique SVPI/SVCI pair between an end-system and the network. Metasignaling messages are transferred on the metasignaling channel identified by VPI = 0 and VCI = 1. Messages that belong to AMTs of the same AP are easily identified, because metasignaling Protocol Data Units (PDUs) are conveyed by single cells. In the metasignaling procedure, a unique signaling channel should be allocated to each AMT of an AP, so that demultiplexing of signaling cells entering the switch through the same AP is straightforward. The SVP/SVC of an AMT may convey signaling cells belonging to different signaling sessions, since these are distinguished by the call reference value in the signaling messages. When an AMT is switched on or enters the coverage area of an AP without prior radio connection to another AP, it associates with the AP, i.e., a radio link between AMT and AP is established. Once the radio link is set up, the AMT broadcasts a request for signaling channel as a part of the REGISTRATION message (figure 5(a)) on the uplink metasignaling channel, which is received by the AP and transparently forwarded to the SWS. In the registration phase (which forms part of the Mobility Manage-

149

(a)

(b)

(c) Figure 5. Call set-up procedures: (a) outgoing call, (b) local call, (c) incoming call.

ment Control), the AMT reports to the SWS its specific address that uniquely identifies the terminal within the SWS area. When the SWS receives the registration request, it obtains an available SVPI/SVCI from its internal database and returns it to the AMT (REGISTERED message), on

150


the downlink broadcast metasignaling channel. The SWS also notifies the AP about the SVPI/SVCI values of the AMT (IDM message of ASCP, see table 2 and figure 5(a)). After successful registration, further signaling between the AMT and the SWS can take place. Part of the registration phase is verification and exchange of security keys for ciphered data transfer. Verification/security issues will not be covered here. In case an AMT is moving from one radio cell to another, it sends a HANDOVER REQUEST message to the SWS requesting a handover to another AP by using the old radio connection. The SWS, upon receipt of a request for handover, obtains a new SVPI/SVCI for the AP that the AMT intends to move to, and returns it to the requesting AMT as a parameter of the handover response message. Subsequently, the signaling and data connections are switched to the new AP. Further signaling, using the newly established SVP/SVC, takes place so as to switch the data connections (VP/VC) or to renegotiate the QoS for ongoing connections. In case the old radio link deteriorates very quickly, or it is abruptly cut off, the AMT communicates directly with the new AP without any prior notification (forward handover). In this case, it is assumed that the handover can only be initiated after the mobile has associated itself with the new AP, and a new signaling channel between the AMT and the SWS (via the new AP) has been established using metasignaling. An alternative scheme to what was presented so far for assigning signaling channels in the wireless ATM system under study assumes that during registration the SVP/SVC pair (that is returned to the mobile terminal with the REGISTERED message) is reserved for each particular AMT within the whole SWS area. All APs are made aware of an SVP/SVC pair assignment through the ASCP. This is an acceptable solution for local-area wireless ATM systems with limited AMT population, but it does not scale well in terms of required AP memory for large installations. The advantage of this approach over the previous one is that there is no need to invoke the Metasignaling entity in the SWS each time the terminal associates with a new AP after the AMT’s registration. Another potential application of the Metasignaling protocol, to satisfy the peculiarities of wireless ATM networks, is the detection of connection loss between an AMT and an AP [35]. When the connection between the AMT and the AP is lost, which can happen in a wireless environment due to several reasons, the ongoing calls are cleared and the reserved resources are freed. To make sure that the signaling connection is valid, the SWS periodically polls the associated AMTs. If there is no response from a particular AMT, the SWS assumes the AMT has been disconnected. Subsequently, the pending connections are released and the SVPI/SVCI is set free for further use. The metasignaling approach follows the philosophy of moving more intelligence and decisions to the AMT and reducing the AP complexity, which in turn yields lower infrastructure costs. In metasignaling, each AMT contains an

extra entity that communicates with the SWS every time a new signaling channel is required, and interacts with the local control entities to inform them about the new SVP/SVC. On the other hand, the AP is kept transparent and simple. Point-to-multipoint signaling, although not used in current applications, is easy to achieve with metasignaling. One multipoint SVC can be assigned to multiple AMTs. Downlink signaling messages can be received by multiple AMTs at the same time.

4. Mobile call control and bearer channel control In the architecture described in section 2, the AN internal operations are transparent to the SWS and the core network. In this framework, it is essential to identify the kind of information maintained in the APs and the SWS and to determine the signaling control information required to be exchanged between them during the establishment, modification, or release of calls. When a mobile initiates a new call (figure 5(a)), its signaling channel transparently conveys a standard call SETUP [23,31] signaling message to the SWS. Upon receipt of this request, the Mobile Call Control (MCC) in the SWS identifies the calling AMT and the called terminal, and contacts the location server to track the location of the calling AMT and the called terminal (if it is mobile). Moreover, MCC instantiates a state machine for the call, creates a record for mapping connections to the call, and performs the necessary routing procedures. An initial call acceptance decision is made, based on the user service profile data and on the QoS requirements set by the AMT [44,45]. In case the request is accepted, the RRM of the AP should be notified of the expected new traffic so that it can decide on the admission in the radio part, and allocate radio resources accordingly. To this end, the traffic parameters of the new connection, or at least a useful subset of them, should be communicated to the AP of the calling AMT. This information gives the opportunity for exercising a policing functionality at the AP, implemented implicitly by its radio bandwidth allocator. It also protects the SWS from the unlikely case where, although the SWS sees or expects availability of radio resources, these are exhausted due to the additional overheads of the MAC layer, or a temporary reduction in radio link quality. The latter is useful in case the CAC of the SWS does not take into account issues specific to the wireless access. Since the CAC decision is taken at the SWS, there is always the opportunity to implement a connection acceptance algorithm customized to the specific access system. Traffic characteristics will appear at the AP together with the QoS requirements, declared as the class of service the specific connection will support. This gives the ability to MAC to implement a set of priorities according to the connection a cell belongs to. To be able to recognize the particular connection class, it is necessary to declare also the VPI/VCI values that will be used.


151

Table 1 ASCP messages. Message

Direction

Parameters

ALLOC ALLOC COMPLETE/ALLOC REJECT ALLOC MODIFY RELEASE IDM

SWS to AP AP to SWS AP to SWS SWS to AP SWS to AP Informs the AP about the SVPI/SVCI value of the AMT during registration (see section 2)

MT, MT, MT, MT, MT,

Table 2 Description of ASCP parameters.

4.1. Description of the ASCP The task of the AP-SWS communication and bearer channel establishment in the fixed access network part is undertaken by the ASCP entity. An ALLOC message is generated and forwarded to the AP (figure 5(a)). The AP will reply with an ALLOC COMPLETE or an ALLOC REJECT message indicating whether it agrees or not with the CAC decision. The latter implies that the call is rejected at the SWS. Upon receipt of an ALLOC COMPLETE, the SWS returns a CALL PROCEEDING message to the calling AMT and initiates the call/connection establishment procedures towards the core network (B-ISUP IAM message) if the called terminal is a fixed one [22,31,33]. In case the called terminal is another AMT (i.e., intraSWS call), the call processing module of the SWS triggers the ASCP entity towards the AP of the called terminal, where similar functions to those described above take place (figure 5(b)). The fixed-to-AMT (incoming) call set-up scenario is shown in figure 5(c). The SWS receives an incoming SETUP message, identifies the called AMT, tracks its location, draws an initial CAC decision, and triggers the ASCP entity towards the corresponding AP of the called AMT. In all cases, the ALLOC message transfers to the AP all the connection related information required for the AP operation. This includes the set of parameters shown in tables 1 and 2. An improvement, in case the requested bandwidth or the QoS cannot be supported by the radio part of the communication path, is for the AP RRM to generate an ALLOC MODIFY message indicating this situation and suggesting a QoS degradation needed for the connection to be accepted. This useful “fallback” mechanism intends to set up bearer connections with the highest available bandwidth [51]. However, such a capability in the ASCP protocol is useless if the standard ATM signaling does not support QoS negotiation to let the SWS and the AMT negotiate the new situation. The ASCP is also activated in case the UNI protocol supports modification of the characteristics of already established connections, i.e., upon reception of a MODIFICATION request message. In all scenarios, we have implicitly assumed that AMTs remain stationary at call set-up. If we assume that an AMT may move during call set-up, the set-up might not succeed. In this case, the new location of the AMT is determined

ADR, SVPI, SVCI, VPI, VCI, BWr, CI, CRC MP, CRC MP, VPI, VCI, BWr, CI, CRC VPI, VCI, CRC ADR, SVPI, SVCI, CRC

Parameter

Information

MT ADR SVPI VPI SVCI VCI BWr

Type of the ASCP message The address of the AMT The SVPI value of the signaling connection The VPI value of the data connection The SVCI value of the signaling connection The VCI value of the data connection Bandwidth requested by the connection (e.g., peak, mean and/or sustainable rate) The service class of the particular connection Indication whether the connection is accepted or rejected Cyclic Redundancy Check

CI MP CRC

and another set-up should be attempted following the same procedures. It should be noted that all ASCP messages could be single cell PDUs marked with a pre-agreed VPI activated in all APs upon installation. The message content is error protected via a 32-bit Cyclic Redundancy Code (CRC) specified in the Metasignaling protocol description [21]. A timer is maintained in the ASCP endpoints, which is used to initiate a retransmission procedure in case no reply is received. The calling or called party can initiate the release of the call and its connections. The release procedures in the access network part for the different scenarios are presented in figure 6. Upon receipt of a RELEASE message, the SWS releases all the resources associated with that call and triggers the release of the corresponding connections towards the AP, the core network, or the AMT. 4.2. Support of available bit rate services A potential application of the signaling protocol model described above is the provision of Available Bit Rate (ABR) services to the wireless ATM user [6,24]. The use of ASCP may serve as an efficient mechanism towards this direction. The ABR service support passes through the development of a traffic flow control scheme. The rate-based congestion control implements a feedback mechanism which uses Resource Management (RM) cells for notifying the originator to reduce its rate in case congestion is experienced in any element of the route to the receiver [10].

152


(a)

Figure 7. Backward handover in the wireless ATM access network.

requirement, not necessarily mandatory, is to establish a terminating point for the rate-based control protocol in the AP. That is, upon reception of an RM or ASCP message, the AP reduces the rate at which it transmits ABR traffic to the SWS. This means that the MAC controller allocates permits to ABR cells according to the new bandwidth allocated to the ABR connection. In turn, the AP should be able to issue an RM cell to the AMT with a request to reduce its rate. (b) 5. Handover signaling procedures

(c) Figure 6. Call release procedures: (a) outgoing, (b) local, (c) incoming.

The rate-based congestion control is defined rather as a framework for a family of such mechanisms giving freedom to a number of variances and implementations. It also allows for the application of the mechanism locally; i.e., between any neighboring network elements of the route or in ATM LANs. Bandwidth guarantees for ABR service are defined through the establishment of a minimum rate contract, which can also be used for other near real-time applications. Clearly, ABR service support has an impact on the wireless network design and especially on the MAC. Since the allocated bandwidth changes dynamically, it is necessary to inform the RRM of the AP about the new bandwidth value. Thus, a first requirement is the AP ability to recognize RM cells and to access the bandwidth information carried in their payload. Another approach could use an ASCP message from the SWS to the AP containing the new value of the ABR connection bandwidth. A second

As an active AMT moves between wireless cells, the task of forwarding data between the wired ATM network and the AMT must be transferred to the new cell’s AP through a handover process. In the system considered here, it is assumed that only the mobile terminal monitors the quality of the radio link for the current and the candidate APs, and decides whether to initiate a handover or not. As the focus of this paper is on the ATM signaling control capabilities required for mobility support, we will not discuss the details of the radio aspects of handover. From the signaling viewpoint, two types of handover are identified: backward and forward [37]. Backward handover assumes that the handover is predicted ahead of time and the handover execution is initiated via the old AP. In the forward handover, it is assumed that a new signaling connection is established with the new AP because the AMT disappears abruptly from the old AP’s coverage area, and handover signaling procedures are initiated via the new AP. The backward, intra-SWS handover scenario is depicted in figure 7. When the AMT decides that a handover should be performed, it sends a HANDOVER REQUEST message towards the Mobile Call Control entity of the SWS transparently via the old AP (see section 3). This message contains identification of the AMT, the call and the target AP. An AMT may have multiple active connections at the same time, as multimedia applications are to be supported. If this is the case, during a request for handover the AMT could also indicate the priorities of the different connec-


tions in case the new AP cannot accommodate all the AMT connections. The fast control protocol between the AP and the SWS described in section 4 is also employed here for the release/establishment of the old/new bearers in the fixed network part, and for performing possible QoS renegotiations during handover. Upon receipt of the HANDOVER REQUEST, the SWS identifies the AMT, initiates a state machine for the handover [43], obtains the new VPI/VCI values (and SVPI/SVCI if a dynamic signaling channel allocation scheme is employed), and activates the ASCP entity towards the new AP. Similar procedures to those described in section 4.1 during call set-up are performed between the SWS and the new AP. In this way, the SWS informs the RRM of the target AP about the expected QoS and bandwidth requirements to allocate radio resources accordingly. When the SWS receives the response from the new AP (ALLOC COMPLETE), it sends a HANDOVER RESPONSE message to the AMT to inform it about the handover results, the new VPI/VCI (and SVPI/SVCI) values and possible QoS modifications, and reconfigures the ATM connections towards the new AP. The HANDOVER RESPONSE message indicates to the AMT that it can proceed to perform handover. The AMT releases its radio connection with the old AP, and establishes a radio link with the new AP, retrieving in this way its connectivity with the fixed network. Special ATM (and lower) layer cell relay functions take place at the AMT and the SWS to coordinate the switching of traffic and to guarantee the transport of user data at an agreed QoS level in terms of cell loss, ordering, and delay. Finally, the SWS updates the location server about the new location of the AMT, and sends a RELEASE message to the old AP to notify it that the connection no longer exists and to deallocate the corresponding radio resources. The handover process described above is expected to be quick. Thus, it is not likely that an AMT moves again before the handover is accomplished. But if the AMT moves, handover is again attempted to the current destination AP, and will eventually succeed [39]. The forward handover scenario is similar to the backward one. The AMT releases the old radio connection and communicates directly with the new AP. The main difference with backward handover is that, if a dynamic signaling channel allocation scheme is employed, an SVP/SVC should be obtained prior to any other signaling message exchange (see section 3), since all the signaling messages are sent via the new AP. The handover procedures that take place in the access part of the considered network architecture do not differ substantially in the case of an inter-SWS handover. In an inter-SWS handover, it can be assumed that the old SWS remains as an anchor point for connection rerouting by establishing a data path to the target AP via the fixed Network-to-Network signaling between the old and the new SWS [13,28], similar to that considered within

153

the Private Network-to-Network Interface (PNNI) working group of the ATM Forum [8,39,51]. Additional registration/authentication procedures should take place in this case, considering that the AMT moves to a different domain. In the analysis presented above, it was implicitly assumed that no QoS degradation is observed in the new path (including both radio and fixed connections) after the handover execution. Complications arise if this new path is not able to support the same QoS, or if it cannot accommodate new connections – this situation is mainly considered for the radio part of the path, where bandwidth is a scarce resource. In case the new path supports a lower QoS than the old path, the network (SWS) should be able to start a QoS renegotiation procedure with the AMT that requests a handover. It has to be noted that, given the time constraints imposed by the moving AMTs, QoS re-negotiation may not be feasible [43]. Thus, handover of one or more connections may fail due to lack of radio resources, but the connections for which sufficient resources exist can still be handed over. The ASCP defined in section 4 fully supports these requirements. In case the requested bandwidth or the QoS cannot be supported by the radio part of the new AP, its RRM generates an ALLOC MODIFY message indicating this situation and suggesting the QoS degradation needed for the connection to be accepted. Taking into account the real-time features of the ASCP and the fact that the RRM immediately generates a request for QoS modification, coping with such situations imposes no extra delay on handover. Clearly, if satisfactory QoS cannot be obtained at the new AP, then the handover is rejected. 6. Performance model and numerical results 6.1. Performance model description In this section, the performance of the proposed access signaling protocol design is quantitatively studied via simulation. The objective is to evaluate the mean call set-up and mean handover signaling processing delay, and to study the behavior and the capacity of the access system under light and heavy signaling loads. A comparative performance analysis with an alternate signaling access architecture is also carried out. For the simulation, we consider an access network configuration that consists of 20 APs attached to one SWS, as shown in the high-level system configuration of figure 2. For this configuration, the following assumptions are made [45]: • The moving AMTs are uniformly spread within the access system, so that the number of users within each AP area is the same. • No handovers occur during call or bearer set-up. • Generation of call set-up/release or handover requests is considered as Poisson.

154

N.H. Loukas et al. / A signaling architecture for wireless ATM access networks Table 3 Processing times of the performance model. Parameter Mobile Call Control (MCC)

ASCP RRM Location Server/Data Base SAAL processing time at AP (for “Full Q.2931”) Signaling operations at the AMT

Processing time 30 20 10 5 5 15 5 1

ms ms ms ms ms ms ms ms

(call set-up operation) (handover operation) (call release operation) (for all operations) (for all operations) (invocation processing time) (update processing time) (for all operations)

2.5 ms (for all operations)

arate controllers [9,50]. The service times for each block and for the different message types are shown in table 3, taken from [9,34,44,45,48–50]. Interblock system delays between the subsystems within a network node are taken to be equal to 5 ms in analogy to [44,50]. On each signaling link between access network nodes, we ignore propagation delays, transmission times, and link queuing delays. In the signaling system under study, we assume that 20% of the total call set-up requests that arrive at the SWS comprise outgoing calls (from AMTs to the core ATM/B-ISDN), another 20% refers to incoming calls (from the core ATM/B-ISDN towards AMTs), while the remaining 60% represents local calls (between AMTs under the same SWS area). The duration of each call is exponentially distributed with mean duration time equal to 3 minutes [45]. We consider two scenarios concerning user mobility:

Figure 8. Signaling processing performance model.

• The moving AMTs do not move again before the handover is accomplished. • The same SVP/SVC pair is reserved for each particular AMT within the whole SWS area during registration. In order to develop a convenient model for performance evaluation, we assume that protocol processing for signaling activity does not involve concurrent processing of messages on the same connection [19]. According to this assumption, the typical workload in signaling protocols is considered primarily as a sequence of short tasks determined by the protocol specification, and the interaction between protocol entities is mostly of the request–response type. Figure 8 illustrates the processing model that represents the signaling processing activity, taking into account the signaling procedures that were described in detail in sections 3–5. In this figure, each block represents a functional group (or a subsystem in the terminology of [9]), that is, the model describes the signaling procedures activated in the SWS and the AP. The signaling processing model considers only “Layer 3” signaling messages processing [51]. Processing delays introduced by the lower layers are not taken into account. We assume that all functional blocks are executed by sep-

• Scenario 1. Low mobility model. • Scenario 2. High mobility model. In the first scenario, we assume that 50% of the AMTs are moving. A moving AMT is idle (not in a call) or active (involved in a call). Idle AMTs simply register their location as they move. The SWS maintains AMT location information in the location server/database system. The high mobility model scenario assumes that all AMTs (100%) are moving within the SWS area. This second scenario is used to obtain the maximum signaling processing limits of the proposed signaling architecture. The mean rate of handover or location update requests for each moving AMT is 1 handover request/min (for an active AMT), or 1 location updating request/min (for an idle AMT), respectively. The system is tested by gradually increasing the number of AMTs per AP and consequently the total number of AMTs in the same SWS area, assuming each time various (outgoing, incoming and local) call arrival rates per AMT for both (low and high) mobility scenarios. The simulation tool used for this performance evaluation is the OPNET 2.5A. Our analysis focuses on three performance measures. First, the mean delay from the start of a call SETUP message to the receipt of the CALL PROCEEDING message at the calling AMT (for outgoing and local calls) or the SWS (for incoming calls) is measured. This delay is referred to as Tcp . The CALL PROCEEDING message acknowledges


the SETUP message, and indicates that the call is being processed and (according to the existing standards) no more call establishment information will be accepted [31]. The second performance measure is the mean end-to-end delay (Tset-up ) to establish local calls, between two AMTs. The expiration interval of the corresponding timers, defined by the existing standards [23], serves as the upper bound limit for the call set-up processing delay in the access network part. The third measure is the delay from the start of a HANDOVER REQUEST until the receipt of HANDOVER RESPONSE at the AMT (Tresp ), which is the critical time interval for the MCC at the SWS to decide on handover acceptance and to establish the new bearers in the fixed radio part.

155

In case a dynamic signaling channel allocation scheme (see section 3) is employed in the system under study, an extra delay should be added to the overall handover processing time. Assuming that the processing delay in the Metasignaling entity at the SWS is considered to be analogous to the cell size in octets (an assumption followed in all rele-

6.2. Numerical results Figures 9 and 10 illustrate how the call set-up delay measures Tcp and Tset-up increase with the call arrival rate per AMT for the low and high mobility scenarios, respectively. Note the large increase of the call set-up delay in the high mobility scenario. From these figures, we can obtain the maximum call arrival rate that the system can support in each case. Figure 11 plots the call set-up delay versus call arrival rate for the two mobility scenarios. Note the sizable difference between the maximum acceptable call arrival rates corresponding to the two scenarios. This comes from the fact that in the high mobility scenario the signaling processing entities of the system are used many times during an AMT call’s lifetime due to the very high rate of handover requests. Clearly, this effect is substantially lower when the percentage of mobile users is reduced to 50%. Figures 12 and 13 illustrate analogous results for the mean handover processing delay, Tresp . From the results shown in figures 9–13, we can obtain the maximum number of terminals that the system can accommodate depending on the user mobility and the call arrival rate. We should note here that the presence of lower layers in a real wireless ATM access system will further increase the call set-up and handover processing delays.

Figure 9. Mean call set-up delay for low mobility scenario.

Figure 10. Mean call set-up delay for high mobility scenario.

Figure 11. Mean call set-up delay vs call arrival rate for the two mobility scenarios.

Figure 12. Mean handover delay for the low mobility scenario.

156


Figure 14. Protocol stack at the AP for the “Full Q.2931” system.

Figure 13. Mean handover delay for the high mobility scenario.

vant studies for estimating protocol processing times), then the processing delay for each Metasignaling message could be taken to be equal to half of the 5 ms value (this is the value suggested in [9] for the processing delay of an 90octet Initial Address Message (IAM) of the ISDN User Part (ISUP) protocol) [48]. Under this assumption, we expect that the impact of the Metasignaling (in case of forward handover) would add a delay of no more than 3 msec to the handover processing delay, which does not affect significantly the obtained results. This is the maximum delay observed at the Metasignaling processing entity assuming the maximum number of users shown in the simulation results, the high mobility scenario, and that all the handover requests are issued via the new AP (forward handover), which is the worst-case scenario for moving users. The second set of simulation results compares the proposed signaling architecture with an alternative signaling protocol structure, similar to that considered in other wireless and mobile ATM systems [41,51]. This configuration assumes that the APs realize the full UNI signaling protocol stack (figure 14), and in this way the signaling procedures are performed sequentially from AMT to AP and from AP to SWS (and vice versa). We refer to this kind of structure as the “Full Q.2931” approach. The protocol information flows for call/connection set-up and handover for this architecture are shown in figure 15. Moreover, it is assumed that the SWS is responsible for the location management within its coverage area. The objective of this comparison is to examine the signaling performance differences of the two architectures in terms of Tcp , Tset-up and Tresp . The low mobility model is employed, while the mean call arrival rate is 10 calls/hr/AMT. The Signaling ATM Adaptation Layer (SAAL) processing delay at the AP has been taken into account for the “Full Q.2931” architecture. The processing parameters shown in table 3 have been used. Figures 16 and 17 give the obtained results for the mean call set-up and handover processing delays, respectively, showing the performance advantage of the proposed signaling architecture.

(a)

(b) Figure 15. Generic signaling procedures for the “Full Q.2931” architecture: (a) call establishment, (b) handover (path rerouting scheme [2]).

The performance difference between the two architectures is due to the call/bearer control separation adopted in our approach, referred to as “ASCP”, which decouples the call control from bearer channel control operations. This results in reduced connection control processing and simplifies the handover control (there is no need for creating a call control association at the new AP-SWS interface in case of handover). In the “ASCP” approach, the call processing application (as seen from the UNI viewpoint) is broken into the mobile call control application process and the channel control application process, allowing the distributed (in the


Figure 16. Mean call set-up delays for the two architectures.

Figure 17. Mean handover delays for the two architectures.

SWS and the AP) and parallel execution of the channel control operations along with the call control operations. This justifies the use of reduced processing times for the channel control procedures of the “ASCP” system, and is in agreement with the design choices and the results shown in [44,45], which elaborate on a distributed call processing architecture for broadband signaling systems. On the other hand, the “Full Q.2931” approach executes a monolithic call processing application at both the AP and the SWS, where the overall call process is performed in atomic steps without the use of any modular blocks or parallelism. In this context, the signaling protocol operations at the AP and the SWS can be considered equivalent (i.e., equal signaling protocol service times) resulting in higher total mean processing delays for signaling. In addition, during handover, a new call control association has to be created between the SWS and the new AP, making the overall handover process slower. A performance improvement in the case of the “Full Q.2931” architecture can be achieved, if a less complex mobile call control signaling application is employed at the AP level compared to that executed in the SWS. This consideration is in accordance with the wireless ATM AN design

157

Figure 18. Signaling performance of the “Full Q.2931” architecture with reduced application processing times for MCC.

guidelines, given in sections 1 and 2, and justifies the use of lower signaling processing times for the signaling operations at the AP than those reported in table 3 and used in the previous experiments. Under this assumption, the establishment (release) of the new (old) connection(s) towards the new (old) AP, in case of handover, can be considered as a lightweight operation [27], with reduced processing times also at the SWS. The simulation results shown in figure 18 illustrate how the reduced MCC processing times1 improve the call control and handover performance. Observe from figures 16–18 that the “Full Q.2931” system exhibits similar performance to that of the “ASCP” system when the lower values for the MCC operations are used (20 ms for call set-up, 10 ms for handover, 5 ms for release).

7. Conclusions In this paper, a broadband, multiservice, wireless ATM access system was explored from a signaling protocol architecture viewpoint. The signaling access structure considered extends the broadband V interface concept to wireless ATM access systems. An enhanced version of the VB5.2 control protocol is employed to enable the dynamic operation of the access network and serve the access network internal functions. It has been shown that the use of the proposed access signaling architecture provides cost effective implementations, and simplifies the call/bearer and handover control in the wireless ATM access network part. The obtained results capture the effect of the architectural and access signaling protocol design on the performance of call and mobility control in terms of call set-up and handover mean processing delay. Finally, a performance comparison of our approach with an alternative access signaling configuration has been performed. 1

At the AP for all operations and at the SWS only for handover-related connection handling.

158


8. List of abbreviations AAL ABR ACM AN ANM AP ATM AMT ARQ ASCP B-BCCP BCC(E) B-ISDN B-ISUP BW CAC CC CI CRC ETSI FDD IAM IDM ISUP ITU LAN LAIP MAC MCC(E) MM(E) MT PCS PDU PHY PNNI PON QoS RCC(E) (R)RM SAAL SN SVC(I) SVP(I) SWS TDD TDMA UNI UPC VC(I) VP(I)

ATM Adaptation Layer Available Bit Rate B-ISUP message: Address Complete Access Network B-ISUP message: Answer Access Point Asynchronous Transfer Mode ATM Mobile Terminal Automatic Repeat Query AP-SWS Control Protocol Broadband Bearer Channel Control Protocol Bearer Channel Control (Entity) Broadband-Integrated Services Digital Network Broadband ISDN User Part Bandwidth (ASCP message parameter) Call/Connection Admission Control Call Control Class Indication (ASCP message parameter) Cyclic Redundancy Check (ASCP message parameter) European Telecommunications Standard Institute Frequency Division Duplex B-ISUP message: Initial Address Message ID Mapping (ASCP message) ISDN User Part International Telecommunications Union Local Area Network Local exchange Access network Interaction Protocol Medium Access Control Mobile Call Control (Entity) Mobility Management (Entity) Message Type (ASCP message parameter) Personal Communications Network Protocol Data Unit Physical Layer Private Network-to-Network Interface Passive Optical Network Quality of Service Radio Channel Control (Entity) (Radio) Resource Management Signaling AAL Service Node Signaling Virtual Channel (Identifier) Signaling Virtual Path (Identifier) Switch-Work Station Time Division Duplex Time Division Multiple Access User–Network Interface Usage Parameter Control Virtual Channel (Identifier) Virtual Path (Identifier)

References [1] A. Acampora and M. Naghshineh, An architecture and methodology for mobile-executed handoff in cellular ATM networks, IEEE Journal on Selected Areas in Communications 12 (1994) 1365–1375. [2] A. Acharya et al., Mobility management in wireless ATM networks, IEEE Communication Magazine 35 (1997) 100–109. [3] P. Agrawal et al., SWAN: A mobile multimedia wireless network, IEEE Personal Communication Magazine 3 (1996) 18–33. [4] P. Agrawal, P. Mishra and M.B. Srivastava, Network architecture for mobile and wireless ATM, in: Proc. IEEE 16th Int. Conf. on Distributed Computing Systems (1996) pp. 299–310. [5] B.A. Akyol and D.C. Cox, Signaling alternatives in a wireless ATM network, IEEE Journal on Selected Areas in Communications 15 (1997) 35–49. [6] ATM Forum, Traffic management specification, version 4 (1995). [7] ATM Forum, UNI specification, version 4 (1995). [8] ATM Forum, Private network–network interface specification, PNNI version 1.0 (1996). [9] M. Bafutto et al., Capacity and performance analysis of signaling networks in multivendor environments, IEEE Journal on Selected Areas in Communications 12 (1994) 490–500. [10] F. Bonomi and K.W. Fendick, The rate-based flow control framework for available bit rate ATM service, IEEE Network Magazine 9 (1995) 25–39. [11] CCITT Recommendation I.121, Broadband aspects of ISDN (1988). [12] K. Chen, MAC of wireless LANs for mobile computing, IEEE Network Magazine 8 (1994) 50–63. [13] W. Chen, Impact of anchor re-routing based inter-switch handoffs in wireless ATM access networks, in: Proc. IEEE ICC ’96 (1996) pp. 235–239. [14] F.-C. Cheng and J.M. Holtzman, Wireless intelligent ATM network and protocol design for future personal communication systems, IEEE Journal on Selected Areas in Communications 15 (1997) 1289– 1307. [15] J.H. Condon et al., A wireless ATM local area network using infrared links, in: Proc. ACM Mobicom ’95 (1995) pp. 151–160. [16] K.Y. Eng et al., A wireless broadband ad-hoc ATM local area network, Wireless Networks 1 (1995) 161–174. [17] ETSI Draft Report DTR/SPS-03040, Signaling protocols and switching – V interfaces at the digital service node, version 1.5 (1996). [18] ETSI Draft Report DTR/TM2222, The management of access networks. [19] D. Ghosal et al., Parallel architectures for processing high speed network signaling protocols, IEEE/ACM Transactions on Networking 3 (1995) 716–728. [20] G.A. Halls, HIPERLAN: The high performance radio local area network standard, Electric & Communication Engineering Journal (1994) 289–296. [21] ITU-T Recommendation Q.2120, B-ISDN metasignaling protocol (1993). [22] ITU-T Recommendation Q.2761, Broadband integrated services digital network user part (1993). [23] ITU-T Recommendation Q.2931, B-ISDN user network interface layer 3 specification for basic call/bearer control (1993). [24] ITU-T Recommendation I.371, Traffic control and congestion in B-ISDN (1995). [25] U. Killat, Access to B-ISDN via PONs: ATM Communication in Practice (Wiley-Teubner, 1996). [26] I. Leslie and D. McAuley, Fairisle: An ATM network for the local area, ACM Computer Communication Review 21 (1991). [27] Li et al., Unified handoff control protocol for dynamic path rerouting in mobile ATM networks, in: Proc. IEEE PIMRC ’98 (1998).

N.H. Loukas et al. / A signaling architecture for wireless ATM access networks [28] N.H. Loukas et al., Access signaling and IN-based mobility management for wireless ATM systems, in: Proc. IEEE ICUPC ’98 (1998) pp. 863–867. [29] L.F. Merakos and F. Li, Integrating data traffic in enhanced-TDMA digital cellular systems, in: Proc. Int. Zurich Seminar on Mobile Communications ’94 (1994). [30] J. Mikkonen et al., The magic WAND – functional overview, IEEE Journal on Selected Areas in Communications 16 (1998) 953–972. [31] D. Minoli and G. Dobrowski, Principles of Signaling for Cell Relay and Frame Relay (Artech House, 1995). [32] P. Mishra and M.B. Srivastava, Effect of connections rerouting on application performance in mobile networks, IEEE Transactions on Computers 47 (1998) 371–390. [33] A. Montaressi and R. Skoog, An overview of SS7, IEEE Proceedings 80 (1992) 590–606. [34] D. Niehaus et al., Performance benchmarking of signaling in ATM networks, IEEE Communication Magazine 35 (1997) 134–143. [35] N. Passas et al., Alternatives of signaling channel allocation in wireless ATM networks, in: Proc. of Int. Workshop on Mobile Communications ’96 (1996) pp. 149–153. [36] N. Passas, N.H. Loukas and L. Merakos, A scheduling technique for bandwidth allocation in wireless personal communication networks, International Journal of Wireless Information Networks 4 (1997) 55– 61. [37] G.P. Pollini et al., Signaling traffic volume generated by mobile and personal communications, IEEE Communication Magazine 33 (1995) 60–65. [38] J. Porter and A. Hopper, An overview of the ORL wireless ATM system, in: IEEE ATM Workshop ’95 (1995). [39] B. Rajagopalan, Mobility management in integrated wireless-ATM networks, in: Proc. ACM Mobicom ’95 (1995) pp. 127–141. [40] D. Raychaudhuri and N.D. Wilson, ATM-based transport architecture for multiservices wireless personal communication networks, IEEE Journal on Selected Areas in Communications 12 (1994) 1401–1414. [41] D. Raychaudhuri et al., WATMnet: A prototype wireless ATM system for multimedia personal communication, IEEE Journal on Selected Areas in Communications 15 (1997) 83–95. [42] I. Rubin and S. Shambayati, Performance evaluation of a reservation random access scheme for packetized wireless systems with call control and hand-off loading, ACM Wireless Journal 1 (1995) 147– 160. [43] C.K. Toh, A hybrid handover protocol for local area wireless ATM networks, Mobile Networks and Applications 1 (1996) 313–334. [44] M. Veeraraghavan et al., An alternative approach to call/connection control in broadband switching systems, IEEE Communication Magazine 33 (1995) 90–96. [45] M. Veeraraghavan et al., A distributed strategy for wireless ATM networks, Wireless Networks 1 (1995) 323–339. [46] M. Veeraraghavan et al., Mobility and connection management in a wireless ATM LAN, IEEE Journal on Selected Areas in Communications 15 (1997) 50–68. [47] I.S. Venieris et al., Service management and connection identification in shared medium access networks, in: Network Management and Control, eds. I.T. Frisch, M. Malek and S.S. Panwar (Plenum Press, New York, 1994). [48] I.S. Venieris et al., Architectural and control aspects of the multi-host ATM subscriber loop, Journal of Network and Systems Management 5 (1997) 55–71. [49] J.-K. Wey et al., Traffic impacts of international roaming on mobile and personal communications with distributed data management, Mobile Networks and Applications 3 (1998) 345–356. [50] G. Willmann and P.J. Kuehn, Performance modeling of SS7, IEEE Communication Magazine 28 (1990) 44–56. [51] R. Yuan et al., A signaling and control architecture for mobility support in wireless ATM networks, in: Proc. IEEE ICC ’96 (1996) pp. 478–484.

159

Nikos H. Loukas received the B.Sc. degree with distinction (first in his class) in electrical and electronic engineering in 1992 from the Hellenic Air Force Academy, Greece, and the M.Sc. degree in networking and telecommunications engineering in 1996 from the University of Athens, Greece. Currently, he is a Ph.D. candidate in the same institution in the area of wireless networking. He has participated in several European Union projects. His research interests include mobile networks, signalling protocols, wired and wireless ATM networks and intelligent networks. He is member of IEEE and ACM. E-mail: [email protected]

Nikos Passas received the Diploma in computer engineering from the University of Patras, Greece, and the Ph.D. degree from the University of Athens, in 1992 and 1997, respectively. From 1992 to 1995 he was with the Greek National Research Center “Demokritos”, working as a Systems and Network Administrator in the National Academic and Research Network “Ariadne”. From 1995 to 1997 he worked as a senior researcher in the ACTS project “Magic WAND”. His research interests are in the protocol design and performance analysis for mobile networks, and multimedia communications. E-mail: [email protected]

Lazaros Merakos received his Diploma in electrical and mechanical engineering from the National Technical University of Athens, Greece in 1978, his M.Sc. and Ph.D. degrees in electrical engineering from the State University of New York, Buffalo, in 1981 and 1984, respectively. From 1983 to 1986, he was on the Faculty of Electrical Engineering and Computer Science at the University of Connecticut, Storrs. From 1986 to 1994 he was on the Faculty of Electrical and Computer Engineering Department at Northeastern University, Boston, MA. During the period 1993–1994 he served as a director of the Communications and Digital Signal Processing (CDSP) Center at Northeastern University. During the summers of 1990 and 1991 he was a Visiting Scientist at the IBM T.J. Watson Research Center Yorktown Heights, N.Y. In 1994 he joined the Faculty of the University of Athens, Greece, where he is presently a Professor in the Department of Informatics. His research interests include mobile networks, high speed networks and multimedia communications. E-mail: [email protected]

Iakovos S. Venieris received the Dipl.-Ing. degree from the University of Patras, Patras, Greece, in 1988, and the Ph.D. degree from the National Technical University of Athens (NTUA), Athens, Greece, in 1990, all in electrical & computer engineering. He is currently an Assistant Professor in the Electrical and Computer Engineering Department of NTUA. His research interests are in the fields of B-ISDN, high speed switches, Internet, internetworking, signalling, and performance evaluation. He has several publications in the above areas. Dr. Venieris has received several national and international awards for academic achievement. He has been exposed to standardisation body work and has contributed to ETSI and ITU-T. He has participated in several European Union and national projects. He is member of IEEE and the Technical Chamber of Greece. E-mail: [email protected]