Low-power Access Protocols Based on Scheduling for Wireless and Mobile ATM Networks Krishna M. Sivalingam1 , Mani B. Srivastava2 , Prathima Agrawal3 , Jyh-Cheng Chen4 1
School of Electrical Engg. & Computer Science, Washington State University, Pullman, WA 99164 2 Dept. of Electrical Engg, UCLA, Los Angeles, CA 90095 3 AT&T Labs, Whippany, NJ 07981 4 Dept. of Electrical & Computer Engg., State University of New York at Buffalo, Buffalo, NY 14260
Abstract—
This paper presents the design and analysis of a medium access control protocol called EC-MAC (Energy Conserving Medium Access Protocol) that supports multimedia traffic for wireless ATM networks. The objective of protocol design is to develop a low-power access protocol that will provide support for different traffic types with quality-of-service (QoS). The network architecture is derived from a testbed built at Bell Labs called SWAN (Seamless Wireless ATM network). The network is based on the infrastructure model where one base station serves all the mobiles currently in its cell. A reservation based approach is proposed, with appropriate scheduling of the requests from the mobiles. This strategy is utilized to accomplish the goals of reduced power consumption and support service quality provision in wireless links. Simulation based performance analysis of the protocol for voice, video and data traffic is presented. I. I NTRODUCTION Wireless services, such as cellular voice, PCS (Personal Communication Services), mobile data, and wireless LANs, are some of the strongest growth areas in telecommunications today [1]. Third-generation networks designed to carry multimedia traffic such as voice, video, audio, animation, images, and data transmission are under intensive research investigation. The goal of the wireless networking research is to provide seamless communications, high bandwidth, and guaranteed quality-of-service regardless of location and mobility constraints. The bandwidth offered by the wireless network will typically tend to lag behind that offered by the wired network. The wired network will serve as the primary or backbone system with enormous bandwidth, while the wireless network will extend the reach of the network. In order to avoid a serious mismatch between future wired and wireless networks, broadband wireless systems should offer similar services as the current and proposed wired broadband networks. These wired broadband systems, such as BISDN ATM, are expected to offer constant bit-rate (CBR), variable bit-rate (VBR), and available bit-rate (ABR) services designed to support multimedia applications. Wired ATM networks are characterized by support for diverse traffic types including multimedia. ATM networks are based on a cell-based transport mechanism with the switching provided by high performance hardware. The source and destination establish bidirectional virtual-circuits (VC) for communication and the ATM network provides sequential delivery of cells. Each VC in the wired ATM network is provided quality-of-service (QoS) guarantees, implemented using call admission control, traffic shaping 0 The authors can be reached at
[email protected],
[email protected],
[email protected],
[email protected].
and policing, and cell scheduling mechanisms. Research efforts are currently underway to study the integration of wireless and ATM networks. Results from these early efforts have been reported in [2–6]. The objective of integrating ATM and wireless networks is to ensure that the the wired ATM network services are seamlessly extended to mobile and wireless users. One of the fundamental challenges in extending the ATM network to the wireless domain is to extend the virtual circuit with quality-of-service to mobile connections. We believe that support for this functionality needs to be provided at the wireless medium access layer. Traditional access protocols for wireless networks do not consider quality-of-service issues or diverse traffic types as envisioned for multimedia networks. To provide CBR, VBR and ABR services to end users, a wireless MAC protocol must be able to provide bandwidth on demand with different levels of service. The network could offer different levels of service with mobile applications adapting to the offered service quality as required. The objective of this paper is to present the design and analysis of a medium access protocol, referred to as EC-MAC (Energy Conserving Medium Access Protocol). The protocol design is driven by two major factors. The first is that the access protocol should be energy-efficient since the mobiles typically have limited battery power. The second factor is that the protocol should provide support for multiple traffic types, with appropriate quality-of-service levels for each type. The dual goals of low power consumption and QoS provision lead us to a protocol which is based on reservation and scheduling strategies. Sharing of the wireless channel among multiple mobiles and connections requires that some form of statistical multiplexing be used. The base station (BS) receives transmission requests or VC-setup requests from the mobiles. The BS schedules the time slots on the channels to the mobiles based on this information. The key to providing service quality will be the scheduling algorithm executed at the base station. The new features of the protocol design described in this paper are the consideration of low-power operation, multiple traffic types, and service quality with respect to offered bandwidth. II. N ETWORK A RCHITECTURE The network architecture is derived from the SWAN network built at Bell Labs [3] – one of the first wireless ATM network testbeds. Figure 1 shows the functional blocks in the wireless last hop. The primary function of the basestation is to switch cells among various wired and wireless ATM adapters attached to the basestation under the control of a Connection Manager signaling module. The basestation is effectively an ATM switch that has wireless (RF) ATM adapters on some of its ports. At the other end of the wireless last hop is the mobile that has a RF wireless adapter, a connection signaling manager module, and a module that routes cells from/to various software/hardware agents acting as sinks and sources of ATM cells within the mobile. The connection managers
ATM Cell Router Adapter Interfaces
ATM Cells AAL Packets
ATM Cells ATM Cell Router Adapter Interfaces
ATM Cells Wired ATM Adapter
ATM Cells
Reservation New Control User Sched. Phase Phase Beacon (TDM) (Aloha)
Reduce Collisions
RF ATM Adapter
RF ATM Adapter
Link Cells Air Interface Packets
Reduce Receiver OnTime
BaseStation to Mobile
Reduce TurnAround
Mobile to BaseStation
BS BEACON
ATM Cells
Connection Manager
BS BEACON
TRANSMISSION FRAME Applications
Connection Manager
Reduce Collisions & TurnAround
Fig. 2. Definition of the different phases in EC-MAC protocol.
Fig. 1. The Wireless Last Hop.
at the mobile and the basestations implement, among other things, the VC rerouting protocols required to handle mobility. The subset of the wireless last hop that is of relevance in making ATM wireless is the shaded area in the picture – streams of ATM cells belonging to different VCs, with different QoS requirements, from the higher level ATM layers need to be multiplexed across the wireless link between a mobile and its basestation. Whereas in the wired scenario each host has a dedicated point-to-point link to the corresponding switch port, the wireless case requires the air resources to be shared among the various mobiles located in a cell and communicating to the switch (that is, the basestation). The need to take QoS into account necessitates a unification of the normally disjoint functions of cell multiplexing and medium access control. At the time of initial VC set-up, the connection manager in the ATM layer contacts the MAC subsystem to perform an admission control check based on the quality of service requirements of the connection. Similarly, following a hand-off, the connection manager at a basestation contacts the MAC subsystem to verify if the quality of service requirements of the newly routed VC can be supported. III. P ROPOSED P ROTOCOL This section describes the traffic types the protocol is initially designed for, the design decisions involved, and the proposed ECMAC access protocol. Traffic types: The recently completed ATM Forum Traffic Management (TM) 4.0 specifications, identifies five service categories: (i) CBR – Constant bit rate, (ii) rt-VBR – Real-Time variable bit rate, (iii) nrt-VBR – Non-real-time variable bit rate, (iv) UBR – Unspecified bit rate and (v) ABR – Available bit rate. The proposed protocol considers three types of traffic: CBR, VBR, and UBR. Low-power issues: Mobile computers typically have limited energy for computing and communications because of the short battery lifetimes. Conserving battery power in mobiles should be a crucial consideration in designing protocols for mobile computing. This issue should be considered through all layers of the protocol stack, including the application layer. Typically, low-power design is done at the hardware layers. The protocol design in this paper is directed toward power conservation in link layer and access layer protocol activities. The chief sources of power consumption from the access protocol perspective include the radio transceiver and the CPU. Maximum power is consumed in transmit mode, followed by lower consumption in receive mode and standby mode. For example, Lucent’s Wavelan 2.4 Ghz radio requires 1.725W in transmit, 1.475W in receive, and 0.08W in standby modes for 2 Mbps and
15dBm. In addition, turnaround between transmit and receive modes (and vice-versa) typically takes between 6 to 30 microseconds. The protocols should be defined such that power consumption due to the transceiver and CPU is low, and turnaround is minimized. Possible techniques to minimize power consumption at the MAC level are described in [7]. A. Protocol definition As described earlier, one of the objectives of extending ATM network services to wireless networks is to provide end-to-end service quality. In wired networks, QoS is achieved using appropriate scheduling algorithms at the ATM switches. A similar principle has to be applied to the wireless network where the basestation is an extended ATM switch with wireless and mobility support. The general consensus observed in recent research on wireless ATM networks is that some form of reservation combined with a scheduling mechanism should be provided at the MAC level [8]. Consistent with this idea, the protocol we propose is based on using a scheduling algorithm to allocate bandwidth to the VCs. The access protocol is defined for an infra-structure network with a single basestation serving mobiles in its coverage area. This definition is extensible to an ad-hoc network by letting the mobiles elect a coordinator to perform the functions of the basestation. Each registered mobile is represented by a unique MACid which may be reassigned after handoff to a new basestation. Each VC in every mobile is represented by a VCid which is unique within the mobile. The pair represents a unique VC within a mobile. This is similar to the combination of IP addresses and port numbers in TCP/IP networks. The 24-bit VPI/VCI used in wired ATM networks can be compressed into a unique VCid of smaller length by the basestation. Transmission in EC-MAC is organized by the basestation into frames. Each frame is composed of a fixed number of slots, where each slot equals the basic unit of wireless data transmission. The basic unit of transmission in the wired ATM network is defined to be 53 bytes. In the wireless network, a different slot size may be used. For example, the SWAN testbed uses a slot-size of 64 bytes which arises due to current hardware limitations. The extra bytes may be used for wireless link header and control information. The frame is divided into multiple phases as shown in Fig. 2. Frame Beacon: At the start of each frame, the BS transmits the beacon on the downlink. The beacon contains framing and synchronization information, the uplink transmission order for reservations, and the number of slots in the new user phase. Reservation Phase: The reservation phase is composed of uplink request transmissions from the mobiles. To conserve battery power, the reservation phase should preferably not operate in a contention mode. The reasons to avoid collisions have been outlined in [7]. The reservation phase is made collision-less by letting the basestation broadcast a list containing the set of the mobile IDs and the transmission order. Each mobile is allocated one slot during the reservation phase. Studies in the performance anal-
b0
b1
0 0 1
0 1 0
MACid Rec. MobileID Send. MobileID Send. MobileID
VCid VCid VCid VCid
Description Downlink VC Uplink Peer-to-peer/multicast
Table 1. Packet types.
ysis section show that the maximum number of mobiles that can be supported using a single channel is of the order of a few tens of mobiles. Therefore, the overhead incurred in broadcasting this identifier list during the startup beacon is of the order of tens of bytes. During the uplink phase, each registered mobile transmits (i) new connection requests, (ii) queue status of established VBR and UBR queues. In case of CBR traffic, a specified number of slots are reserved in every frame for a specific VC until the mobile indicates that it is done with the VC transmission. New-User Phase: This phase allows new mobiles to register with the basestation. This phase is operated in a contention mode, using Slotted Aloha. The length of this phase is variable. The basestation beacon broadcasts the available number of slots for user registration during this phase. The basestation initially starts with a small number of slots, and dynamically varies the number of slots based on collisions exceeding threshold values. The basestation transmits all the acknowledgments and registration information for each mobile in the subsequent beacon. Schedule Beacon: The basestation broadcasts a beacon that contains the slot allocations for the subsequent data phase. The data phase includes downlink transmissions from the basestation, and uplink transmissions from the mobiles. Each permission consists of a 2-bit type field, the VCid field and the length field. The allocation information is grouped based on sender ID with a length field. This approach is attractive from a power consumption perspective. It reduces the time the receiver has to be turned on to receive the schedule information. Table 1 lists the different types of permissions. As wireless speeds increase to 20 Mbps and above, equalization is required at the receiver to establish bit synchronization. This synchronization may require around 400-500 bits for each transmission, since each slot may potentially originate from a different sender. Under such conditions, it will not be efficient to make allocations based on individual 53-byte cells. Therefore, our proposed protocol allocates clusters of slots to a sender. Downlink and Uplink Data Phases: Downlink transmission from the basestation to the mobiles is scheduled considering the QoS requirements of the individual VCs. The uplink slots are allocated for CBR, VBR, and UBR traffic. A scheduling algorithm is used to allocate the slots to the traffic sources, and is described below. The definition of the protocol in terms of multiple phases in a frame is similar to other protocols proposed earlier [6, 8–10]. The new features of the proposed protocol are support for multiple traffic types, provision of per-VC queuing and scheduling, low power consideration, and provision of bandwidth service quality to the connections. B. Packet Scheduling Downlink transmission is for cells that have to be forwarded to the mobiles. The downlink transmission schedule considers the individual QoS requirements of the virtual circuits. The scheduling algorithm is similar to the uplink algorithm. It is important to consider the effects of turnaround and receiver turn-on times in the
scheduling of this phase. To minimize this overhead, transmissions to a mobile are done in contiguous slots. The data phase for the uplink transmission from the mobiles is divided into three parts, one each for CBR, VBR, and UBR traffic. The number of slots allocated to the categories is changed dynamically depending on traffic characteristics. The basestation maintains a request table to record the current queue status, and number of slots reserved for each VC in the previous frame. Mobiles update the queue status in each reservation phase. The simulation results presented in Section IV consider three types of traffic – one each for CBR, VBR, and UBR category. Voice is modeled as a two-phase process with talkspurts and silent gaps [11, 12]. Typically, such modeling classifies voice as VBR. We consider that the voice source generates a continuous bitstream during talkspurts and is therefore classified as a CBR source in our scheduling. Video is considered as an example of a VBR source with variable number of cells per frame. Data such as ftp, http and email is considered as an example of UBR traffic. The basestation first schedules slots in the CBR compartment for CBR sources with backlogged queues. The number of slots alloted per CBR source is dependent on the source’s QoS parameters. After allocating slots to existing CBR sources, the basestation schedules other available slots in CBR compartment for newly accepted CBR VCs. Once a CBR source gets permission to transmit, the basestation updates the request table to reserve slot(s) for that mobile. Since VBR sources generate variable number of cells, the basestation needs to dynamically allocate slots for each such source. The basestation first checks to see how many slots have been reserved in the previous frame for a given VC. Based on the new queue status, the basestation reduces the number of slots reserved for each VBR source. This could happen if more slots were reserved than were generated by a VBR video source. After this, the basestation evenly distributes available slots to VBR sources with backlogged queues. Similarly, the basestation evenly decreases the number of slots reserved for each video source if the total reservation is larger than the total number of slots in VBR compartment. The slots allocated to a VC are contiguous to reduce turnaround overhead. After scheduling CBR and VBR sources, the basestation checks for available slots in CBR and VBR compartments. If slots are available, basestation schedules them to UBR sources sharing the slots equally among the backlogged sources. The scheduling algorithm described above is fairly simple since our initial objective is to gain an understanding of the protocol performance. More sophisticated algorithms have been studied in related fields such as ATM switch scheduling. We are currently studying these algorithms to analyze their suitability for the wireless environment. IV. P ERFORMANCE A NALYSIS This section describes simulation results for EC-MAC using realistic source traffic models for video, voice, and data services. The performance of the protocol has been studied through discreteevent simulation. Simulation results have been obtained using stochastic self-driven discrete-event models, written in C with YACSIM package [13]. A. Source Models In simulation, each mobile is capable of generating three different types of traffic: data, voice, and video. An idle mobile generates new data messages, voice calls, and video calls, and all the arrival processes are modeled as Poisson. To focus on the performance of the access mechanism, the channel is assumed to be ideal such that there is no distortion, noise, or other interference for packet transmissions.
0.1
frame generates variable number of ATM cells sent in 3 TDMA frames. On average, each video source reserves approximately 35 slots per frame. Video cells are dropped if not transmitted after 144 ms which is equivalent to 12 TDMA frames. Data is considered as an example of UBR traffic. The length of a data message is exponentially distributed with mean of 5.12 KBytes. The data queue is infinite, and a new data cell is appended to the end of the queue.
Voice Cell Dropped Rate
0.09
[CBR=4;VBR=137;UBR=1]
0.08
[CBR=5;VBR=135;UBR=2]
0.07
[CBR=6;VBR=133;UBR=3] 0.06 0.05 0.04 0.03 0.02 0.01
(a)
0 1
2
3
4
5
6
7
8
Average Data Cell Delay (seconds)
10000
[CBR=4;VBR=137;UBR=1]
1000
[CBR=5;VBR=135;UBR=2]
(10 176 12) 53
[CBR=6;VBR=133;UBR=3]
100 10 1 0.1 0.01 0.001
(b)
1
2
3
4
5
6
7
8
5
6
7
8
1
Video Cell Dropped Rate
0.9
[CBR=4;VBR=137;UBR=1]
0.8
[CBR=5;VBR=135;UBR=2] 0.7
[CBR=6;VBR=133;UBR=3]
0.6 0.5 0.4 0.3 0.2 0.1 0 1
(c)
2
3
4
B. System parameters varied We performed initial simulations for a channel rate of 10 Mbps (10,176 Kbps). A TDMA frame length of 12 ms is considered, and one ATM cell can be transmitted in each TDMA slot. With 10,176 ; = bytes slots Kbps channel rate, there are per MAC frame. The maximum number of mobiles supported is set to 8, with each mobile generating three 3 different traffic types. However, each mobile can have at most one active session of a given traffic type, at any instant. Thus, there are 24 transmission orders downlink and 24 requests uplink in the reservation phase. Each voice and video source gets one permission from schedule beacon, while each data source gets at most 3 permissions if the system load is light. Therefore, the maximum number of permissions for eight mobiles is 40. There are 211 bytes in reservation phase and schedule beacon, transmitted in 4 slots. Therefore, 284 slots are available for data phase, of which 142 slots each are allocated to the downlink and uplink. The system is analyzed under heavy traffic. We set the load of each traffic type source as 99%. Assuming an average length of a voice call as 300 seconds and 99% load, the average time between calls is calculated to be approximately 3 seconds. The average length of a video call is 19.2 minutes and the average time between calls is calculated to be approximately 11 seconds.
= 288
1 0.9
Channel Utilization
0.8 0.7 0.6 0.5 0.4
[CBR=4;VBR=137;UBR=1]
0.3
[CBR=5;VBR=135;UBR=2]
0.2
[CBR=6;VBR=133;UBR=3]
0.1 0 1
(d)
2
3
4
5
6
7
8
Number of Mobiles
Fig. 3. Performance of EC-MAC for a network rate of 10 Mbps and each mobile with 99% load.
A voice source is modeled as a two-state Markov process with 32 Kbps coding rate representing a source with a slow speech activity detector (SAD) [11]. Measured values for talkspurts and silence are 1.00 sec and 1.35 sec [11], each with exponential distritalkspurts and sibution. This results in an average of lence gaps for each voice conversation. A voice cell is dropped if not transmitted after 36 ms. When a new voice cell arrives at a full queue, the first cell in the voice queue will be dropped. Video is modeled as a VBR source and is generated using a 19.2 minute video conference as in [12]. The number of ATM cells per video frame is described by a gamma distribution and a DAR (1) (first-order discrete autoregressive process) model determined by three parameters: mean, variance, and correlation. We chose a modified version of the H.261 video coding standard with mean rate of 104.8 ATM cells and peak rate of 220 cells per video frame, and standard deviation of 29.7. The TDMA frame length is taken to be 12ms, and the video interframe rate is 36 ms which equals 27.78 video frames per second. Based on these parameters, a video
36%
64%
C. Performance metrics studied The focus of the study is to understand the service quality provided by the protocol with an increase in the number of mobiles supported. To this end, we define the following QoS parameters: Voice-cell dropped rate is defined as the number of voice cells dropped per total number of voice cells generated. One suggested value is that it should not exceed 1% [9] , otherwise the distortion is perceptible. Data cell delay is defined as the time between generation and transmission completion. Video-cell dropped rate is defined as the number of video cells dropped per total number of video cells generated. The desired value depends on the coding algorithm and the type of service. Channel utilization is defined as the number of slots used for transmission per total number of slots available in a frame. All the QoS parameters are with respect to the uplink only. Downlink parameters will be considered in further study. D. Simulation Results Fig. 3 shows the results for 10 Mbps channel rate and each moload. In the figures, the representation [CBR=4; bile with VBR=137; UBR=1] means there are 4 slots, 137 slots, and 1 slot in the CBR, VBR, and UBR compartment, respectively. This is abbreviated to [4;137;1] in the text below. The objective of the study is to identify the maximum number of mobiles that can be accommodated for a desired QoS. Voice-cell dropped rate is considered first. As expected, less voice cells are dropped if more slots are reserved for voice traffic. Fig. 3(a) indicates that six mobiles can be supported within voice cell dropped rate with [4;137;1], seven mobiles with the [5;135;2], and more than 8 mobiles with [6;133;3]. When the number of mobiles is 7, voice cell dropped rate is around 0.03 for [4;137;1], 0.01 for [5;135;2], 0 for [6;133;3]. After adding one more mobile, voice cell dropped rate increases quickly to 0.1 for [4;137;1], to 0.02 for [5;135;2], and to 0.004 for [6;133;3]. This
99%
1%
Energy (Watt-second)
0.1
cases. The analysis details and comparison to a number of other access protocols may be found in [14].
0.01
0.001
802.11 EC-MAC (Min) EC-MAC (Max) 0.0001 5
10
15
20
Number of Mobiles: G = 0.50 Fig. 4. Power consumption (energy in watt-second) for transmitting a single packet using Lucent’s 15 dBm 2.4 GHz radio.
indicates that a slight increase in reservation for voice traffic substantially increases voice quality. Fig. 3(b) compares the data cell delay for the different reservations. Since data is transmitted with lower priority, the data cell delay increases very fast when the number of mobiles increases. Since video traffic takes many more slots than other service types, there are many slots available for data traffic in VBR compartment = when the number of video sources is less than 3.8 ( ). After this point, data traffic can use only slot(s) in UBR or CBR compartment. We can see that the data cell delay is less than 0.01 sec when the number of mobiles is less than 4 for [6;133;3]. The data cell delay saturates when the number of mobiles is approximately 5 for [4;137;1], 6 for [5;135;2], and 7 for [6;133;3]. Fig. 3(c) shows that video cell drop rate increases rapidly after the number of mobiles is 3 regardless of the number of slots reserved for VBR. Since the number of cells generated by video sources is much higher than the number of cells generated by voice and data sources, a small reduction in number of cells in VBR compartment have very little affect on video quality. However, it does improve the quality for voice and data traffic. In order to improve the video quality, a good admission control scheme which limits the number of VBR video sources in the system is necessary. Fig. 3(d) examines uplink channel utilization. When more slots are reserved for video, overall utilization is high but there is little performance difference due to variations in the allocation of slots. The channel utilization increases as the number of mobiles increases, and saturates when number of mobiles is approximately 4. Fig. 4 provides a comparison of power consumption for ECMAC and IEEE 802.11 standard while transmitting a single packet using Lucent’s radio described in Section II. In figure, G is the offered traffic load which includes newly generated plus retransmitted packets. The value plotted on the y-axis is determined using the time the transmitter and receiver have to be turned on, to transmit a single packet. Since 802.11 is based on carrier sensing, the receiver is expected to be continously monitoring the medium while transmission. This is expensive from a power consumption perspective. EC-MAC attempts to alleviate this problem through reservation and scheduling. The receiver is turned on only to receive the acknowledgment and the schedule which indicates the transmission slots. Depending on what time the mobile gets its schedule beacon, the receiver usage time in EC-MAC may range from one slot to the maximum number of slots in schedule beacon. These are represented as the upper and lower bounds for EC-MAC. The figure shows power consumption is less than 802.11 in both
35+1)
= 142 (1 +
V. S UMMARY This paper describes EC-MAC, an access protocol for wireless and mobile ATM networks. The goals of the access protocol are to conserve battery power, to support multiple traffic classes, and to provide different levels of service quality for bandwidth allocation. The protocol is based on a combination of reservation and scheduling mechanisms. An analysis of the various QoS parameters based on number of mobiles in the cell is presented. For instance, a voice cell drop rate of less than 1% can be sustained for more than eight mobiles for a single channel. A more detailed analysis of the performance with other scheduling algorithms is under study. The reduced power consumption characteristics of EC-MAC with respect to IEEE 802.11 is also presented in the paper. R EFERENCES [1] K. Pahlavan and A. H. Levesque, “Wireless data communications,” Proceedings of the IEEE, vol. 82, pp. 1398–1430, Sept. 1994. [2] J. Porter and A. Hopper, “An ATM based protocol for Wireless LANs,” Tech. Rep. TR-94-2, Olivetti Research Labs, UK, Apr. 1994. [3] P. Agrawal, E. Hyden, P. Krzyzanowski, P. Mishra, M. Srivastava, and J. A. Trotter, “SWAN: A Mobile Multimedia wireless network,” IEEE Personal Communications, pp. 18–33, Apr. 1996. [4] M. Naghshineh (Guest Ed.), “Wireless ATM: Special Issue.” IEEE Personal Communications, Aug. 1996. [5] Z. Liu, Medium Access Control Schemes for DS-CDMA Wireless Packet Networks. PhD thesis, University of Pennsylvania, December 1995. [6] D. Raychaudhuri, L. J. French, R. J. Siracusa, S. K. Biswas, R. Yuan, P. Narasimhan, and C. A. Johnston, “WATMnet: A Prototype Wireless ATM System for Multimedia Personal Communication ,” IEEE Journal on Selected Areas in Communications, vol. 15, pp. 83–95, Jan. 1997. [7] K. M. Sivalingam, M. B. Srivastava, and P. Agrawal, “Low power link and access protocols for wireless multimedia networks,” in IEEE VTC, (Phoenix, AZ), May 1997. [8] G. Bianchi, F. Borgonovo, L. Fratta, L. Musumeci, and M. Zorzi, “C-PRMA: A centralized packet reservation multiple access for local wireless communications,” IEEE Transactions on Vehicular Technology, vol. 46, pp. 422–436, May 1997. [9] D. J. Goodman, R. A. Valenzuela, K. T. Gayliard, and B. Ramamurthi, “Packet reservation multiple access for local wireless communications,” IEEE Transactions on Communications, vol. COM37, no. 8, pp. 885–890, 1989. [10] K. S. Natarajan, “A Hybrid Medium Access Control Protocol for Wireless LANs,” in Proc. IEEE Intl. Conf. on Selected Topics in Wireless Communications, 1992. [11] D. J. Goodman and S. X. Wei, “Efficiency of packet reservation multiple access,” IEEE Transactions on Vehicular Technology, vol. 40, pp. 170–176, Feb. 1991. [12] D. Heyman, T. V. Lakshman, A. Tabatabai, and H. Heeke, “Modeling teleconference traffic from VBR video coders,” in International Conference on Communications (ICC), (New Orleans, LA, USA), pp. 1744–1748, May 1994. [13] J. R. Jump, YACSIM Reference Manual. Rice University, Department of Electrical and Computer Engineering, 1.2 ed., August 1992. [14] P. Agrawal, J.-C. Chen, and K. M. Sivalingam, “Battery power consumption based analysis of MAC protocols for wireless multimedia networks,” tech. rep., Washington State University, Pullman, WA, July 1997.
