Wireless Networks 7, 159–171, 2001 2001 Kluwer Academic Publishers. Manufactured in The Netherlands.A Capacity Analysis for the IEEE 802.11 MAC ProtocolY.C. TAY and K.C. CHUANational University of SingaporeAbstract. The IEEE 802.11 MAC protocol provides shared access to a wireless channel. This paper uses an analytic model to study thechannel capacity – i.e., maximum throughput – when using the basic access (two-way handshaking) method in this protocol. It providesclosed-form approximations for the probability of collision p, the maximum throughput S and the limit on the number of stations in awireless cell.The analysis also shows that: p does not depend on the packet length, the latency in crossing the MAC and physical layers, theacknowledgment timeout, the interframe spaces and the slot size; p and S (and other performance measures) depend on the minimumwindow size W and the number of stations n only through a gap g = W/(n − 1) – consequently, halving W is like doubling n;themaximum contention window size has minimal effect on p and S; the choice of W that maximizes S is proportional to the square rootof the packet length; S is maximum when transmission rate (including collisions) equals the reciprocal of transmission time, and thishappens when channel wastage due to collisions balances idle bandwidth caused by backoffs.The results suggest guidelines on when and how W can be adjusted to suit measured traffic, thus making the protocol adaptive.Keywords: IEEE 802.11 MAC protocol, capacity analysis, saturation throughput, closed-form approximation, analytic validation, win-dow size adaptation1. IntroductionWith the proliferation of mobile computers, their limitedcomputing resources, and the popularity of Internet access,there is a growing need for these computers to be networked.In response, the IEEE 802.11 study group proposed a stan-dard for wireless local area networks[8]. This standardspec-ifies the characteristics of the physical layer, as well as themedium access control (MAC) protocols in the link layer.There are essentially two MAC protocols in the proposal–abasic access method that uses two-way handshaking(DATA-ACK) and a RTS/CTS variant that uses request-to-send and clear-to-send messages in a four-way handshake(RTS-CTS-DATA-ACK). This paper analyzes the former butnot the latter, for two reasons: (1) the basic access method ismandatory, whereas RTS/CTS is an optional variant; (2) theperformancefor RTS/CTS is significantly differentfrom thatfor basic access [2], and therefore requires a separate an-alytic model. We also do not discuss the no-ACK optionmeant for broadcasts and multicasts, nor the point coordina-tion function, which is an optional polling scheme definedon top of basic access.Basic access uses carrier-sensing multiple access withcollision avoidance (CSMA/CA). There are numerousCSMA protocols, and their performance under low load con-ditionsare usuallysimilar [22,24]. The many variationsarisebecause of efforts to improve on performance and push backthe limits. Our analysis therefore focuses on the most im-portant such limit – namely, the maximum (or saturation)throughput, which measures the capacity when the protocolis used to access the channel, and which is lower than theraw bandwidth for the physical medium itself.Our model considers the case where multiple stations usethe protocol to share a wireless channel without a coordinat-ing base station. It assumes that the stations are homoge-neous in traffic generation, channel noise is negligible, andthere are no hidden terminals. A scenario that may fit theseassumptions would be a classroom or meeting in which stu-dents or executives exchange information on their laptops.From the modeling perspective, it is not difficult (but some-what tedious) to take noise into consideration; also, hiddenterminals require a separate model and, in any case, shouldbe analyzed together with RTS/CTS because the two areclosely related [2,3,7].In contrast to previous simulation studies of the 802.11MAC protocols [2,13,23], the performance analysis wepresent here is based on a mathematical model. This modelnot only differs from previous analytic models of the 802.11protocols [3,6,7,10], it is also different from the othertechniques in the CSMA literature [1,5,12,14,15,17–20].Whereas these studies use stochastic analysis (e.g., Markovchains), our model uses the average value for a variablewherever possible – this is a methodology that is commonlyused in the performance analysis of computer systems. Thistechnique is simple, yet effective: It provides closed-formexpressions for the probability of a collision and the satu-ration throughput, thus facilitating the analysis of variousissues, such as the choice of window size, the limit on thenumber of stations, and the tradeoff between collisions andbackoffs. It also yields two rules of thumb: halving the ini-tial window size W (for the exponentialbackoff) is similar ineffect to doubling the number of stations, and the optimumchoice of W is proportional to the square root of packet size.A performance model is usually validated by comparingits numerical predictions with simulation results. For our160 TAY AND CHUAmodel, we also check its analytical conclusions against sim-ulated performance – we call this analytic validation.Weuse Bianchi et al. simulator [2] for the validation. This sim-ulator is comprehensive in capturing the many details in the802.11 protocol, and although our model omits many ofthese details and relies on many approximations, the com-parison shows that it is accurate both numerically and ana-lytically.We begin in section 2 by describing the protocol, and in-troduce the performance model in section 3. We then checkthe numerical accuracy of the model in section 4, before us-ing it to analyze the protocol in section 5; there, we con-stantly use the simulator to check the results from the analy-sis. Section 6 summarizes our conclusions.2. Protocol descriptionThe basic access method for the 802.11 MAC protocolworks as follows: To send a packet, a station X first listensto the channel for time TDIFS(DIFS is distributed interframespace). If there is silence for TDIFS, X proceeds with thetransmission (e.g., station A in figure 1); otherwise, X waitsfor the first TDIFSof silence after the current busy period,then backs off for a random interval (e.g., station C in fig-ure 1).For each packet, X initializes a contentionwindowsizeeWto be W ,theminimum
View Full Document
Unlocking...