1111
`
`■
`
`■
`
`Editor-in-Chief:
`lmrich Chlamtac
`
`i
`
`Volume 1 (1996) No. 3
`ISSN 1383 469X
`Published December 1996
`
`

`

`Mobile Networks and Applications
`The journal of special issues on mobility of systems, users, data and computing
`
`Mobile Networks and Applications (MONET) is a joint publication of the ACM and Baltzer Science Publishers
`
`Editor-in-Chief: Imrich Chiamtac, ECS Department, Boston University, 44 Cummington Street, Boston, MA 02215, USA
`E-mail: chlamtacOacm.org
`
`Advisory Board:
`Anthony Acampora, Professor, UC San Diego
`Toby Berger, Professor, Cornell University
`Vinton Cerf. Vice President, MCI
`Domenico Ferrari, Professor, UC Berkeley
`Robert Gallager, Professor, MIT
`Richard Gitlin, Vice President, Lucent Technologies
`David Goodman, Professor, Rutgers University
`
`Area Editors:
`Design and analysis of algorithms, Baruch Awerbuch,
`Johns Hopkins University
`Algorithms for online and mobile environments,
`Maurizio Bonuccelli, University of Pisa, Italy
`Advanced services, multimedia, Jon Crowcroft,
`Univ. College, England
`Systems and technologies, Rene Cruz, UC San Diego
`Wireless, mobile network planning, Beza/el Gavish,
`Vanderbilt University
`Data management and databases, Tomasz Imielinski,
`Rutgers University
`Network/node/host architectures, Raj Jain,
`Ohio State University
`Mobility algorithms, broadband wireless, Ravi Jain, Bellcore
`Mobility and applications, David Johnson, Carnegie Mellon University
`Aims and Scope:
`
`Leonard Kleinrock, Professor, UCLA
`Tom Leighton, Professor, MIT
`Barry M. Leiner, Vice President, MCC
`Ray Miller, Professor, University of Maryland
`Dave Morgan, Vice President, Motorola
`Christos Papadimitriou, Professor, UC Berkeley
`Mischa Schwartz, Professor, Columbia University
`
`Service algorithms and ATM, Mark Karol, Lucent Technologies
`Mobile, roaming services, Randy Katz, UC Berkeley
`Nomadic computing, Arv ind Krishna, IBM
`Security, reliability, availability, Shay Kutten, Technion, Israel
`Standards, Tom La Porta, AT&T
`Network management, Aurel Lazar, Columbia University
`Wireless-wireline systems integration, Lauis Lame, BMDO
`Design of mobile networks, Debasis Mitra, AT&T
`Protocols for mobility, Charlie Perkins, IBM
`Multihop networks, sattelite, PCS, Ambatipudi Sastry, SRI
`Routing and multicasting, Martha Steenstrup, BBN
`Performance characterization, Yutaka Takahasi,
`Nara Institute of Science & Technology, Japan
`Systems for wireless and mobile environments,
`David Tennenhouse, MIT
`
`MO~_ET's technical scope reflects the emerging symbiosis of portable computers and wireless networks, addressing the convergence of
`mob1hty, computing and information organization, its access and management. In approving Special Issues, the journal places an equal emphasis
`on the various areas of nomadic computing, data management, related software and hardware technologies, and mobile user services, alongside
`more "classical" topics in wireless and mobile networking. The journal documents practical and theoretical results which make a fundamental
`contribution, in the following, representative, areas:
`• Nomadic computing, applications and services supporting the mobile user
`• Design and analysis of algorithms for online and mobile environments
`• Protocols to cope with mobility, limited bandwidth, intermittent connectivity
`• Data management issues in mobile environments
`• Mobile and wireless networks and their architectures
`• Mobile node/ host architectures
`• Mobility management, mobile agent and proxy architectures
`• Solutions for portable, mobile and roaming services
`• Mobile applications, location-dependent and sensitive applications
`• Systems and technologies for wireless and mobile environments, wearable computers and body area networks
`• Performance characterization of mobile/ wireless and nomadic systems
`• Design, management and operation of emerging wireless environments
`• Mobile network planning and standardization
`• Integration ofwireline and wireless systems
`• Security, scalability and reliability in wireless communication and computations environments
`• Service algorithms, emerging topics, multimedia, ATM, etc.
`
`Sublcrlption Information
`Subscription Volume 1, 1996: Institutional price: Swiss Francs 354.00 I USS 290.00 including postage.
`Subscription Volume 2. 1997 ( 4 issues): Institutional price: Swiss Francs 365.00 I USS 299.00 including postage.
`Personal pricing on request. In case of exchange rate fluctuations the Swiss Francs Price is definitive.
`Subscriptions should be sent to your usual supplier or to Baltzer Science Publishers, P.O. Box 37208, 1030 AE Amsterdam, The Netherlands
`(aubacribetna.bal tzer.nl). Second class postage paid at Rahway, NJ. Postmaster: please send address corrections to Baltzer Science Publishers, C/O
`Mercury Airfreight International Ltd., 2323 Randolph Avenue, Avenel, NJ 07001, USA. A CM members: MONET is available for its members at the rate of
`USS 47.00; student members: USS 42.00. In the USA please contact: ACM Member Services Department, Church Street Station, P.O. Box 12114, New
`York, NY 10257, USA (aCllbelpeaca.org); members in Europe: ACM European Services Centre, 108 Cowley Road, Oxford OX4 IJF, UK. Phone:
`+44 1865 382338. E-mail: aca-europetaca.org. WWW: http : / /wv.aca.org, http : //wv.acm.org/ j ournals/130.html.
`0 Baltzer Science Publishen (1996): All righta rctCrved. No part of tbil publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic,
`
`mechanical, pbotocopyin1, recording or otherwise, without the prior permission of the publisher. No responsibility ii ass'.""ed by the pu~lilhcr for any injury and/ o~ dama~ to penon.1 :i
`
`property as I matter of producu liability, nesJiacn<e or otherwile, or from any use or operation of any methods, products, tnstructioru or ideas contained 1~ the matenal herein. Although
`advertisins material ii cxpocted to conform to ethical 1tandards, ind111ion in this publication doa 1101 constitute a guarantee or endonement of the quahty or value of such product or of
`the claima made by ill manufacturer. Claims for illuca not received 1bould be made within 3 montba of publication. If not, they cannot be honoured free of charge. Claims, requesta for
`complimentary iuues and req-ts for back iuueaand/orvolll.lllCI 1bould be aent directly to the publisher or to your 111ual supplier.
`ii
`
`

`

`Mobile Networks and Applications
`
`iii
`
`

`

`Mobile Networks and Applications
`The journal of special issues on mobility of systems, users, data and computing
`
`Volume 1 (1996) No. 3
`
`Published December 1996
`
`Contents
`
`Editorial
`
`Wireless A TM - an overview
`G.A. Awater and J. Kruys
`
`Forward error control for MPEG-2 video transport in a wireless A TM LAN
`E. Ayanog/u, P. Pancha, A.R. Reibman and S. Ta/war
`
`A multiple access scheme for multimedia traffic in wireless A TM
`X. Qiu, V. O.K. Li and J.-H. Ju
`
`Mobility management in integrated wireless-A TM networks
`B. Rajagopalan
`
`A signaling and control architecture for mobility support in wireless A TM networks
`R. Yuan, S.K. Biswas, L.J. French, J. Li and D. Raychaudhuri
`Lossless handover for wireless A TM
`H. Mitts, H. Hansen, J. Immonen and S. Veikkolainen
`
`A hybrid handover protocol for local area wireless A TM networks
`C-K Toh
`A proposal of an ATM wireless access system for tetherless multimedia services
`M. Umehira, A. Hashimoto, H. Matsue and M. Nakura
`
`233
`
`235
`
`245
`
`259
`
`273
`
`287
`
`299
`
`313
`
`335
`
`iv
`
`

`

`Mobile Networks and Applications 1 (1996) 259^272
`
`259
`
`A multiple access scheme for multimedia traffic in wireless ATM
`
`Xiaoxin Qiu, Victor O.K. Li and Ji-Her Ju
`
`Communication Sciences Institute, Department of Electrical Engineering, University of Southern California,
`
`Los Angeles, CA 90089-2565, USA
`
`Abstract. This paper proposes a multiple access scheme for the forthcoming wireless ATM (Asynchronous Transfer Mode) sys-
`
`tem. Such ATM compatible wireless systems are motivated by the rapidly increasing demand for wireless extensions to broadband
`
`networks, which are expected to support mixed broadband services including Constant Bit Rate (CBR), Variable Bit Rate (VBR),
`
`and Available Bit Rate (ABR) traffic. Since these different traffics have very different performance requirements, the multiple
`
`access scheme design is very challenging. In this paper, we propose a multiple access scheme called Dynamic Time Division Multiple
`
`Access with Piggybacked Reservation (DTDMA/PR), attempting to achieve higher statistical multiplexing efficiency in the mixed
`
`VBR/CBR/ABR traffic scenario. The basic idea is to exploit two levels of reservation. The first level deals with the isochronous nat-
`
`ure of CBR and VBR traffic and the bursty nature of ABR traffic by using the ALOHA reservation procedure. The second level
`
`exploits the piggybacked reservation approach to cope with the dynamic feature of VBR traffic in order to increase the multiplexing
`
`efficiency. An analytical model is also developed in this paper and verified by simulation. Numerical examples are given to gain
`
`some insight into the protocol itself.
`
`1. Introduction
`
`When the source is active, the CBR packet is generated
`
`periodically. It can tolerate low to medium packet loss
`
`It is generally believed that the wireless system is evol-
`
`but is not very tolerable to delay and delay jitter. On the
`
`ving towards supporting a wide range of telecommunica-
`
`other hand, teleconferencing is a VBR type of service
`
`tion services, including data, voice, video, and images.
`
`with bit rates ranging from 64 Kbps to 384 Kbps. It can
`
`Such ``multimedia capable'' wireless system is motivated
`
`also tolerate packet loss but not delay and delay jitter.
`
`by the rapidly increasing demand for wireless extensions
`
`VBR traffic is also generated periodically but with cer-
`
`to future broadband networks. It is obvious that this
`
`tain dynamic characteristics. The number of packets
`
`extension will be very challenging because the band-
`
`generated in a period is a random variable instead of a
`
`width of the wireless system is much more limited than
`
`fixed number as in CBR. Finally, high-speed data with
`
`that of its wireline counterpart. But it is very important
`
`bit rates from 1 Mbps to 10 Mbps requires high transfer
`
`in the system design to provide at least qualitatively simi-
`
`rate and very low or no packet loss. But it can tolerate
`
`lar attributes even though quantitative equivalence is
`
`some delay and delay jitter [13]. The desired multiple
`
`not feasible. To be compatible with the developing ATM
`
`access scheme must provide mechanisms to deal with
`
`network, the wireless ATM concept is proposed in [13].
`
`each of these service types at reasonable QOS levels.
`
`In that paper, the possibility of supporting broadband
`
`Several multiple access schemes which attempt to
`
`services in the wireless environment is explored from the
`
`cope with the different features of ATM traffic can be
`
`architecture point of view. The same issues are also dis-
`
`found in the literature. Dynamic Time Division Multiple
`
`cussed in [3] from the system viewpoint. Wireless ATM
`
`Access (DTDMA) is discussed in [13] under the inte-
`
`is a very popular topic nowadays, stimulating much
`
`grated voice and data scenario. The frame structure is
`
`research and commercial interests.
`
`shown in Fig. 1(a). Each frame is divided into three peri-
`
`A major technical issue related to the wireless ATM
`
`ods, namely, the reservation subframe, the fixed alloca-
`
`system design which has a significant impact on the user
`
`tion subframe, and the dynamic allocation subframe. In
`
`performance, system capacity and complexity is the
`
`DTDMA, the voice user can make a reservation for a
`
`selection of the appropriate multiple access control
`
`traffic slot in the fixed allocation subframe by using the
`
`scheme. Wireless ATM is expected to support Constant
`
`ALOHA protocol at the beginning of each active period
`
`Bit Rate (CBR) traffic such as voice and CBR video,
`
`(or talkspurt) and can keep the slot until the end of its
`
`Variable Bit Rate (VBR) traffic such as teleconferencing
`
`active period. However, the data user has to make reser-
`
`and dial-up videophone, and Available Bit Rate (ABR)
`
`vations each time he has something to transmit. This
`
`traffic such as packet data, which have very different
`
`protocol is designed to provide the flexibility to accom-
`
`characteristics
`
`(bit-rate
`
`range, Quality-of-Service
`
`modate CBR and ABR traffic. VBR is not considered. A
`
`(QOS) requirement, etc.). For example, voice telephone
`
`similar protocol called TDD ALOHA-Reservation is
`
`traffic is CBR with bit rates ranging from 2.4 Kbps to 32
`
`proposed in [15] for the integration of video (VBR) and
`
`Kbps, depending on the vocoder. It is either active
`
`data (ABR) in wireless LAN. Its frame structure is
`
`(during voice talkspurt) or inactive (between talkspurts).
`
`shown in Fig. 1(b). Each frame is divided into uplink and
`
`Ä J.C. Baltzer AG, Science Publishers
`
`

`

`260
`
`X. Qiu et al. / A multiple access scheme for wireless ATM
`
`I J 111111111 ~ I_
`I
`I Reservation
`
`Minislots
`
`I I
`
`I I
`
`Fixed Allocatioo Subframe
`
`I ~~---J
`
`Subframe
`
`Movable Boundary
`
`(a) DTDMA Protocol Frame Structure
`
`Minislots Minislots
`
`Message Subframe
`
`Message Subframe
`
`Control
`Subframe
`
`Uplink
`
`Downlink
`
`(b) TDD ALOHA-Reservation Protocol Frame Structure
`
`Fig. 1. Frame structures of DTDMA and TDD ALOHA-Reservation protocols.
`
`downlink slots. At the beginning of the uplink slots,
`
`The basic reservation approach designed with CBR
`
`there are several reservation minislots and acknowledg-
`
`(e.g., voice) traffic in mind is not very suitable for VBR
`
`ment minislots for the downlink. A data user with pack-
`
`traffic. For example, how many slots should a VBR user
`
`ets to transmit has to contend for the traffic slots on a
`
`reserve in a frame? Following the same approach as for
`
`frame-by-frame basis by randomly choosing one avail-
`
`the CBR traffic, we can assign slots to the VBR user
`
`able reservation minislot to transmit its reservation
`
`according to either its peak packet arrival rate or its
`
`packet. The video user, on the other hand, will be
`
`mean packet arrival rate or some value between its peak
`
`assigned one of the reservation minislots by the base sta-
`
`and mean rates. The peak-rate assignment guarantees
`
`tion at the connection setup phase and can keep it for its
`
`the user performance but reduces the multiplexing effi-
`
`exclusive use during the lifetime of the connection to
`
`ciency, especially when the ratio of its peak rate to mean
`
`avoid collision and to cope with the dynamic nature of
`
`rate is high. On the other hand, the mean-rate assign-
`
`VBR traffic. In each frame, the video user will tell the
`
`ment increases the system capacity at the expense of per
`
`base station through the dedicated minislot how many
`
`user performance in the form of excessive packet loss if
`
`packets need to be transmitted. The traffic slots will be
`
`no other control mechanisms are applied. The problem
`
`assigned to this user on a frame-by-frame basis accord-
`
`associated with an in-between assignment is that it is
`
`ing to his requirement and the availability of resource.
`
`difficult to determine the optimal number of slots
`
`However, the authors did not discuss how to make the
`
`assigned to the user considering the dynamic nature of
`
`minislot reservations in the first place, although we can
`
`the traffic.
`
`probably assume that it is also through contention with
`
`In this paper, we propose a new multiple access
`
`other users. Of course, using dedicated minislots will
`
`scheme, called Dynamic Time Division Multiple Access
`
`cause more reservation failures for both data users and
`
`with Piggybacked Reservation (DTDMA/PR), bearing
`
`newly arriving VBR users, and therefore degrade the
`
`the above considerations in mind. We exploit the same
`
`system performance. CBR traffic is not considered in
`
`frame structure as that of DTDMA but modify the
`
`that paper. We will extend this protocol to include CBR
`
`reservation procedure in order to deal with the different
`
`traffic and compare it with our new protocol. Besides
`
`characteristics of CBR, VBR, and ABR traffic. Two
`
`these two protocols, several other protocols such as
`
`level reservations are used. In the first level, all the CBR
`
`Dynamic Reservation Multiple Access
`
`[12] and
`
`and VBR users who have just become active and ABR
`
`Distributed-Queueing Request Update Multiple Access
`
`users with packets to transmit have to contend for the
`
`(DQRUMA) [5] have also been proposed to improve the
`
`traffic slots using the ALOHA protocol in the reserva-
`
`multiplexing efficiency under the integrated traffic sce-
`
`tion subframe. At the end of this reservation period, the
`
`nario. However, they focus on integrating voice and
`
`base station will announce the assignment to winners.
`
`data. VBR traffic is not considered. To accommodate
`
`The CBR user will be assigned traffic slots according to
`
`VBR traffic, new mechanisms need to be developed.
`
`its packet arrival rate and will keep using these slots in
`
`

`

`X. Qiu et al. / A multiple access scheme for wireless ATM
`
`261
`
`each subsequent frame as long as it remains active.
`
`2. Dynamic TDMA with Piggybacked Reservation
`
`The VBR user will be assigned a certain number of traf-
`
`(DTDMA/PR) protocol
`
`fic slots (for example, equivalent to its minimum or to
`
`its mean rate) and can also keep the slots until the end
`
`In this section, we describe the DTDMA/PR proto-
`
`of
`
`its current active period. With this reservation
`
`col in detail. For simplicity, a single-cell system is consid-
`
`mechanism, the circuit switched transmission mode is
`
`ered, where the base station and the dispersed users
`
`achieved for CBR and VBR users. On the other hand,
`
`communicate through the uplink (users to base station)
`
`the ABR user will be assigned a number of slots accord-
`
`and the downlink (base station to users) channels. Time
`
`ing to his requirement and the availability of resources
`
`is divided into slots which are grouped into frames. The
`
`and he has to release them at the end of this frame. The
`
`duration of each slot is equivalent to one packet trans-
`
`second level reservation is designed to cope with the
`
`mission time. We assume that all packets have the same
`
`dynamic feature of VBR traffic. Each VBR data packet
`
`packet length. The ABR traffic is bursty. The CBR pack-
`
`has a ``piggyback field'' which can be used to make the
`
`ets are generated periodically during the active period
`
`second level reservation. If the VBR user has more
`
`according to the codec rate. We design the frame length
`
`packets to transmit
`
`in the current frame than its
`
`to coincide with this period. Groups of VBR packets are
`
`assigned slots, it will inform the base station of the
`
`also generated periodically during the active period,
`
`number of additional slots he requires in the piggyback
`
`although the number in each group is a random variable
`
`field. Upon receiving this data packet, extra slots will
`
`following a certain distribution. For the sake of simpli-
`
`be assigned to him if there are still some available after
`
`city, we assume here the periods of VBR and CBR
`
`the requirements of the newly arriving VBR, CBR and
`
`sources are the same. Using layered MPEG coder
`
`ABR traffic have been satisfied. There is no collision in
`
`instead of the standard MPEG coder, different priorities
`
`the second level reservation.
`
`can be assigned to VBR packets [11,14]. The packets
`
`The rest of this paper is organized as follows. In sec-
`
`with higher priority will always be transmitted first to
`
`tion 2, the DTDMA/PR protocol is described in detail.
`
`improve the performance.
`
`Traffic models for CBR/VBR/ABR sources are given in
`
`The uplink frame structure of DTDMA/PR is shown
`
`section 3. The analytical model for the DTDMA/PR
`
`in Fig. 2. Each frame is divided into three parts: the reser-
`
`protocol is developed under the integrated VBR/CBR
`
`vation subframe, the long-term reservable subframe,
`
`traffic scenario in section 4. Several performance meas-
`
`and the short-term reservable subframe. The reservation
`
`ures are defined in the same section. In section 5, we first
`
`subframe consists of a number of reservation minislots.
`
`study the performance of the integrated VBR/CBR sys-
`
`Slots in the long-term reservable subframe can be
`
`tem. The analytical model is verified by simulation.
`
`reserved by CBR or VBR users according to their
`
`Some numerical examples are given in order to gain
`
`requirements. Slots in the short-term reservable sub-
`
`some insight into the protocol itself. The impact of sys-
`
`frame can only be used on a frame-by-frame basis. The
`
`tem and traffic parameters on the system performance is
`
`boundary between these two subframes are movable.
`
`also investigated. Then, the performance of the inte-
`
`The unused long-term reservable slots in the current
`
`grated CBR/VBR/ABR system is simulated and com-
`
`frame can also be used for the short-term reservations.
`
`pared to the existing protocol. Finally, we conclude in
`
`We assume that the base station is powerful enough to
`
`section 6.
`
`handle the downlink traffic.
`
`Resecvation
`Minislots
`
`Long-Term Reservable Subframe
`
`Movable
`Boundary
`
`~ Short-Term
`
`Reservable
`Subframe
`
`ID
`
`No. of Slots
`Required
`
`(CBR or VBR or ABR)
`
`Fig. 2. Frame structures of DTDMA/PR protocol.
`
`No. of Extra Slots
`
`

`

`262
`
`X. Qiu et al. / A multiple access scheme for wireless ATM
`
`When a CBR (VBR) source generates a new active
`
`slots it has already been assigned, there is no packet loss.
`
`period (e.g., a new talkspurt for the voice user), it will
`
`But if it is not so, packet loss may occur. We call those
`
`randomly choose one minislot at the beginning of next
`
`packets the extra packets. Some mechanism has to be
`
`frame and transmit a reservation packet, informing the
`
`designed to handle this situation. We propose the second
`
`base station its traffic type, its Identification Number
`
`level reservation, called ``piggybacked'' reservation. We
`
`(ID), the number of slots required, etc. The active
`
`append a piggyback field to the end of each packet.
`
`ABR user will also transmit the reservation packet in a
`
`Through this field, the base station will be informed of
`
`randomly chosen minislot. At the end of the reserva-
`
`the number of additional slots a VBR user requires in the
`
`tion period, the base station will broadcast the IDs of
`
`current frame. Since only several bits are needed to carry
`
`those users who have made successful reservations, the
`
`this information, the piggyback field has negligible
`
`number of slots assigned to each of them, and the cor-
`
`impact on the packet length. According to the current
`
`responding slot positions. The reservations of VBR
`
`traffic situation, the base station will decide the number
`
`and CBR users can only be made in the long-term
`
`of extra slots assigned to the VBR user in the current
`
`reservable subframe. The ABR sources can take the
`
`frame. Those slots assigned in the piggybacked reserva-
`
`remaining slots which are not reserved by VBR and
`
`tion mode have to be released at the end of the current
`
`CBR users in the current frame. A reservation failure
`
`frame. Note that we assume here the propagation delay
`
`may be caused by either packet collisions in the reser-
`
`and the processing delay at the base station are negligi-
`
`vation period or a shortage of traffic slots. These failed
`
`ble. The VBR user finds its additional assignment imme-
`
`reservations may retry in the next frame. The CBR or
`
`diately through monitoring the downlink channel. This
`
`VBR user who makes the reservation successfully may
`
`assumption makes sense in the micro- or pico-cellular
`
`keep using the assigned slots until the end of its current
`
`environment where the distance between the base station
`
`active period. Afterwards, the slots will be released to
`
`and the user is relatively small. Our performance model
`
`the base station. The ABR user has to release the slots
`
`can be generalized to take the impact of those delays into
`
`after its current transmission. We assume that both the
`
`account.
`
`VBR and the CBR traffic are delay-sensitive and need
`
`It should be noted that to make DTDMA/PR work,
`
`immediate delivery. The packet has to be discarded if
`
`slot reordering and reassignment needs to be implemen-
`
`its delay exceeds a certain limit. Because this packet
`
`ted in order to keep all the unused slots at the end of each
`
`dropping causes performance degradation, it should be
`
`frame. For example, as illustrated in Fig. 3, as CBR user
`
`limited. On the other hand, the ABR traffic is delay-
`
`3 (occupying slot 3 in this example) leaves the system in
`
`insensitive. The ABR packet can be stored in the buffer
`
`frame i at the end of his current ON period, the base sta-
`
`until being successfully transmitted. Note that we
`
`tion will be informed through a special bit pattern in the
`
`assume here all the reservation failures are caused by
`
`last packet of user 3. After the reservation period of
`
`reservation packet collisions or the shortage of traffic
`
`frame i
`
`1, the base station will assign a new user to that
`
`‡
`
`slots. No channel errors are considered. A suitable
`
`newly available slot, or if there is no new users, move the
`
`extension can be made to take the channel
`
`impair-
`
`user currently occupying slot 4 to slot 3. This reordering
`
`ments into account.
`
`and reassignment technique has also been used in
`
`The reservation through the minislot is the first level
`
`Dynamic Channel Assignment (DCA) in TDMA cellu-
`
`reservation which deals with the isochronous nature of
`
`lar systems. See [1,6,16].
`
`CBR and VBR sources and the bursty nature of ABR
`
`sources. It is sufficient for the CBR user since the number
`
`of packets generated in each frame is fixed. However, the
`
`3. Traffic models of CBR/VBR/ABR sources
`
`number of packets generated in each frame by a VBR
`
`source is a random variable. If the number of packets
`
`As the first step towards developing the analytical
`
`generated in the current frame is within the number of
`
`model for DTDMA/PR, we establish the models for
`
`- - occupied - -
`
`, ,
`
`I
`
`11111111 . . . T'i
`
`frame i
`
`I I
`
`I frame i+l
`
`Fig. 3. Illustration of slot recordering and reassignment.
`
`

`

`X. Qiu et al. / A multiple access scheme for wireless ATM
`
`263
`
`traffic sources first. We assume that CBR or VBR traffic
`
`3.2. Traffic model of VBR source
`
`consists of alternating ON (or ACTIVE) and OFF (or
`
`SILENCE) periods. The lengths of ON and OFF periods
`
`ON-OFF sources have been successfully used to
`
`are assumed to be exponentially distributed with average
`
`characterize VBR traffic [4,7]. Taking the teleconferen-
`
`c
`durations t
`
`and t
`
`o
`
`c
`
`s
`
`for CBR sources and tv
`
`and tv
`
`for
`
`cing-type of traffic as an example, there are two typical
`
`o
`
`s
`
`VBR sources. Denote the frame duration as T .
`
`kinds of scenes, the active one and the inactive one [10].
`
`The system can be designed so that when it is inactive, no
`
`3.1. Traffic model of CBR source
`
`packet is generated, which corresponds to the OFF per-
`
`iod. When the user is active, the number of packets gen-
`
`We follow the traffic model developed in [9]. In
`
`erated in each frame is a variable, which distinguishes it
`
`DTDMA/PR protocol, an active CBR user may be in
`
`from the CBR source. This variable can be described by
`
`one of three states, namely, the Silence State (SS), the
`
`its probability distribution function, denoted as
`
`Reservation State (RS), and the Contention State (CS).
`
`Qv
`
`ˆ f
`
`qv
`1 ;
`
`qv
`2; ;
`
`qv
`pv
`
`g
`
`, where qv
`
`i
`
`is the probability of gen-
`
`The RS and CS states belong to the ON period. When
`
`erating i packets in a frame. The maximum number of
`
`the user has nothing to transmit, it is in the silence state.
`
`packets generated by a VBR source in each frame is
`
`When the user generates a new ON period and has not
`
`denoted as pv . It is assumed that there is at least one
`
`made a reservation yet, it leaves the silence state and
`
`packet generated in each frame when the user is active.
`
`enters the contention state. When it contends success-
`
`The VBR user can be in the silence state, the reservation
`
`fully and obtains a reservation, this user will enter the
`
`state, or the contention state. Fig. 4(b) illustrates the
`
`reservation state. Fig. 4(a) illustrates the transitions
`
`transitions among SS, RS, CS states. In this figure, qv
`
`is
`
`s
`
`among these three states. In this figure, q
`
`c
`
`s
`
`is the prob-
`
`the probability that an ON period of VBR source ends in
`
`ability that an ON period of CBR source ends in a frame;
`
`a frame; qv
`
`a
`
`is the probability that an ON period of VBR
`
`c
`
`q
`
`a
`
`is the probability that an ON period of CBR source is
`
`source is generated in a frame; and Pv
`
`is the probability
`
`CR
`
`generated in a frame; and P
`
`c
`
`CR
`
`is the probability that a
`
`that a VBR source obtains a reservation in the current
`
`CBR source obtains a reservation in the current frame.
`
`frame. We assume here the probability that a VBR user
`
`We assume here the probability that a CBR user returns
`
`returns to the silence state before it obtains a reservation
`
`to the silence state before it obtains a reservation is zero.
`
`is zero. Then,
`
`Then,
`
`c
`
`q
`
`a
`
`1
`
`exp
`
`T
`
`ˆ
`
`ÿ
`
`…ÿ
`
`c
`t
`
`†
`
`s
`
`=
`
`and q
`
`c
`
`s
`
`1
`
`ˆ
`
`ÿ
`
`exp
`T
`…ÿ
`
`c
`
`t
`
`†
`
`o
`
`=
`
`1
`
`…
`
`†
`
`qv
`
`a
`
`1
`
`ˆ
`
`ÿ
`
`exp
`…ÿ
`
`T
`
`tv
`
`s
`
`=
`
`†
`
`and qv
`
`s
`
`1
`
`exp
`
`T
`
`ˆ
`
`ÿ
`
`…ÿ
`
`tv
`
`o
`
`=
`
`†
`
`2
`
`…
`
`†
`
`and Pv
`
`CR
`
`also depends on the system parameters. Note
`
`and P
`
`c
`
`CR
`
`depends on the system parameters.
`
`that more sophisticated source model can be used to
`
`·· ...
`
`·••·· .. ~ /\·--... ....
`
`p~
`
`ss
`
`qc
`a
`
`'•·· ...
`
`ON
`
`p•
`CR
`
`•,
`
`··• ...
`
`...
`
`(a) Traffic Model for CBR Source
`
`(b) Traffic Model for VBR Source
`
`Po
`
`Psucc
`
`(c) Traffic Model for ABR Source
`
`Fig. 4. CBR, VBR, and ABR traffic models.
`
`

`

`264
`
`X. Qiu et al. / A multiple access scheme for wireless ATM
`
`describe the VBR traffic. However, no matter what
`
`Simplification. We observe that the CBR (or VBR) user
`
`source model is exploited, the mechanism of DTDMA/
`
`leaving or entering the silence state only depends on
`
`PR remains unchanged.
`
`whether a new ON period is generated or the current one
`
`3.3. Traffic model of ABR source
`
`CBR and VBR) are contending or how many other users
`
`is ended. It does not depend on how many users (both
`
`are in the silence state [8]. Therefore, we can split the
`
`Each ABR user is either in the thinking state (TH) or
`
`four-dimensional Markov chain into three sub-pro-
`
`the backlogged state (BK). When there is no packet gen-
`
`cesses: the active-silence process of CBR, the active-
`
`erated, the user is in the TH state. With a packet gener-
`
`silence process of VBR, and the reservation process
`
`ated, the user enters the BK state immediately and no
`
`which depends on the status of both VBR and CBR
`
`new packet can be generated until the current one has
`
`sources.
`
`been successfully transmitted. The state transitions of
`
`The active-silence process of CBR sources, whose
`
`ABR sources are illustrated in Fig. 4(c), where p
`0
`
`is
`
`state variable is N
`
`c
`
`s
`
`, describes the above transition of
`
`defined as the packet generation rate of ABR users in
`
`CBR users' entering and leaving the silence state. It
`
`terms of packets/frame/ABR user and p
`
`is the trans-
`
`evolves independently of the individual values of N
`
`c
`
`, N
`
`c
`
`,
`
`succ
`
`r
`
`c
`
`mission success probability.
`
`N v
`
`, N v
`
`, and also of N v
`
`r
`
`c
`
`s
`
`. We denote its stationary distribu-
`
`tion as …
`
`c
`
`s
`
`;
`
`† :
`
`c
`
`c
`
`c
`
`c
`
`4. Analytical model for DTDMA/PR
`
`s
`
`…
`
`;
`
`†
`
`
`ˆ …
`
`n
`
`s
`
`
`
`
`
`P
`
`N
`
`†g
`
`ˆ
`
`…
`
`ˆ
`
`s
`
`
`
`n
`
`†
`
`s
`
`:
`
`Similarly, the active-silence process of VBR sources with
`
`For the sake of simplicity, we will concentrate on the
`
`state variable N v
`
`s
`
`can be modeled in the same way. Its sta-
`
`integrated CBR and VBR traffic scenario when we
`
`tionary distribution can be denoted as …
`
`s
`
`;v† :
`
`develop the analytical model in this paper. It is straight-
`
`forward to extend it to the integrated CBR, VBR and
`
`s
`
`;v†
`
`…
`
`
`ˆ …
`
`nv
`
`s
`
`
`
`
`
`P
`
`†g
`
`ˆ
`
`…
`
`N v
`
`s
`
`ˆ
`
`
`
`nv
`
`s
`
`†
`
`:
`
`ABR traffic case but with higher computational com-
`
`On the other hand, the reservation process models
`
`plexity. Therefore, a simulation model will be used to
`
`the reservation or transmission status of the system. It is
`
`obtain the numerical results when we investigate the per-
`
`formance of the integrated VBR/CBR/ABR system.
`
`Denote the total number of traffic slots per frame as
`
`N , the number of long-term reservable slots as N
`
`, the
`
`l
`
`a two-dimensional process whose state variables are
`
`c
`
`N
`
`, N v
`
`f
`
`g
`
`r
`
`r
`
`. Its evolvement depends on the values of both
`
`N
`
`N
`
`c
`
`s
`
`c
`
`s
`
`and N v
`
`s
`
`. Although not completely true with varying
`
`and N v
`
`s
`
`, the stationary distribution of this process is
`
`number of short-term reservab