Simulation of the MAC Portion of IEEE 802.11 and Bursts of Errors for Wireless Data Networks

The focus of this research is to investigate the effects of bursts of errors and packet collisions on the performance of the medium access control (MAC) portion of the IEEE 802.11 wireless local area network (LAN) protocol.An important ingredient in rapid expansion of wireless networks is the seamless transition between wired and wireless systems. The IEEE standards group in charge of developing the widely used IEEE 802.3 LAN standard has developed the IEEE 802.11 wireless LAN standard. IEEE 802.11 remains hidden from the upper levels of the network, thus allowing a seamless transition between networks. The foundation protocol for the IEEE 802.11 standard, known as Distributed Foundation Wireless Medium Access Control (DFWMAC), operates at the MAC level of the Data Link Layer. The protocol bases its access control mechanism on a principle called Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA), which is an adaptation of the Carrier Sense Multiple Access with Collision Detection (CSMA/CD) protocol used by IEEE 802.3 standard. The collision avoidance scheme in CSMA/CA allows data packets to be transferred via the wireless medium with lower probability of packet collision. In a slotted multi-access wireless system, performance parameters are affected by the bit error rates on the communication channel. These errors occur as a result of noise introduced by the radio channel or data packet collisions. Collisions occur when two or more stations select the same time slot to transmit their data, thus causing corruption in data packets. In this research, a simulation model coded in Microsoft's Visual Basic programming environment is utilized to investigate the effects of bit errors and packet collisions on performance in CSMA/CA. Performance parameters used in this study include throughput, medium utilization, collisions and station data queue lengths. In the simulation model, error bursts in the communication channel are modeled using a simple Gilbert model with two states, good (G) and bad (B). State G is error free, thus errors can only occur while the model is in state B. Collisions are simulated by two or more stations starting to transmit data packets in the same time slot. Therefore, as the number of stations increases, more and more stations compete for the medium, resulting in an increase in the number of collisions. Collisions are also increased by the amount of traffic that each station introduces into the system. Station load is defined here as the number of data packets per unit time that are released by the higher network protocol layers.The results in Chapter 5 demonstrate that higher network throughput can be achieved when the aggregate load on the network is distributed. For example, 30 stations offering 20 kilobits per second (kbps) of load for a total of 600 kpbs, results in a network throughput of 585 kbps. However, three stations offering 200 kbps of load for a total of 600 kbps offered load, results in a network throughput of 486 kbps. The distributed load is serviced at a 17 percent higher rate. However, once the network becomes saturated at above 40 stations for this model, collisions will more than offset the performance gains produced by the distribution of load.Furthermore, reducing the packet size by 50 percent under an approximately 19.5 percent packet error rate results in a 12 percent gain in throughput. This is primarily due to higher utilization of the network by shorter packets. However, as the packet error rate is reduced, the performance gap between the two packet sizes is reduced. Once the errors are removed completely from the communications channel, the longer packets produce a higher throughput than the shorter packets.