Design and performance study of a media access control protocol for wireless lans

Abstract

Wireless local area networks are gaining importance for computer communications since they eliminate the problem of laying interconnecting LAN cables within and across buildings. One of the main issues in wireless LANs is the optimal sharing of a common wireless medium by multiple users. Media Access Control (MAC) protocols play a vital role in fair distribution of the channel resource to all nodes and aim at achieving maximum throughput performance with minimum delay. This dissertation deals with the design and performance analysis of a new protocol for media access in a multiuser Wireless LAN environment. The objectives of this work include: • Design of a new multiaccess protocol suitable for wireless LANs. • Estimation of throughput and delay performance of the protocol under various network models involving different arrival processes, capture effect, finite buffer capacity, dual channel operation, high speed and multihop environments through Markov chain analysis and simulation. The proposed Optimum Channel Utilisation Multiaccess (OCUM) protocol is hybrid in design, conceptually based on slotted ALOHA and busy tone sensing techniques. It uses two independent wireless channels namely Message Channel (M-Channel) and Busy tone Channel (B-Channel). The M-Channel is a slotted, high bandwidth channel used for data transmission and the B-Channel is meant for carrying a narrow band busy tone to indicate the activity on the M-Channel. The OCUM protocol adopts unique features like preamble transmit policy at the source node prior to the transmission of actual data, busy tone broadcast policy initiated at the destination node and transmission of variable length messages (measured in number of data packets). The OCUM protocol results in efficient utilisation of the M-Channel and overcomes the problem of hidden nodes in multihop environments. The throughput performance of OCUM protocol in a fully connected wireless LAN with instantaneous busy tone feedback and no buffer conditions is studied through the analysis of a finite-state and discrete-time Markov chain. Analytical expressions are derived for the M-Channel idle rate, collision rate and throughput. Numerical results show that the protocol offers good throughput performance with maximum channel utilisation of the order of 0.784. Channel utilisation is further improved for larger message sizes, making the OCUM protocol suitable for file transfer type of applications. The protocol shows unstable behaviour at high arrival rates like any other random access protocol. Next, the effect of capture on the throughput performance of OCUM protocol is estimated through Markov chain analysis. We consider intentional capture realised through three different Multiple Transmit Power Level (MTPL) schemes. MTPL Schemes I & II are two-level schemes and MTPL Scheme III is an m-level scheme. Numerical results show throughputs around 0.9 in all the three capture schemes. It is shown that a capture threshold, Ct = 3 achieves throughput performance close to infinite capture threshold in Schemes I & II. In Scheme III, the throughput improvement tends to get saturated for m > 6. The performance of the OCUM protocol is further studied with finite buffer capacity (L) at each node and a message reject policy that discards new messages arriving at nodes when the message buffers are full. The throughput and delay performance of the protocol in a fully connected network is estimated through simulation. Results show that the protocol offers constant throughput of the order of 0.8 at high arrival rates irrespective of the buffer capacity. At low arrival rates, the throughput improves as the value of L is increased. However, the improvement becomes marginal for L > 4. For large message sizes (> 50 packets), close to 100% channel utilisation is realised at the cost of increased delay performance. We also study the Dual Channel operation of OCUM protocol (DC-OCUM) which uses two pairs of message and busy tone channels. The throughput performance of the DC-OCUM protocol has been estimated for fully connected network with finite buffers through simulation. Two types of channel selection policies for making a transmission attempt are studied and compared. In the first channel select policy (sense and pick), sensing of both B-Channels is performed first and an idle M-Channel is picked for transmission. In the second policy (pick and sense), any one of the M-Channels is picked first and the corresponding B-Channel is sensed for making a transmission attempt. It is shown that the former policy performs better than the latter. In high speed networks, the propagation and processing delays may exceed the packet duration which violates the instantaneous busy tone feedback assumption. The OCUM protocol has been modified to operate in such high speed environments. The throughput performance of the Modified OCUM (MOCUM) protocol is estimated through simulation. The protocol is found to offer maximum throughputs of the order of 0.4 – 0.7 for wireless LANs up to 100 Mbps data rate over 10 Kms range. Finally, the throughput and delay performance of the OCUM protocol in multihop environments is studied. Partially connected static topologies with finite number of nodes that are globally synchronised to the slot duration are considered. An Input Buffer Limiting (IBL) policy that rejects entry of new messages into a node when the number of queued up messages exceeds a specific threshold has been adopted. Simulation results for ring, bus, multi-connected ring, square lattice and other arbitrary topologies are presented. The throughput performance improves beyond unity due to spatial reuse in multihop operation.