Scheduling in multiservice wireless networks

Licentiatavhandling, 2006

Wireless access networks that support services such as voice, video telephony, email and web-surfing are being deployed world-wide today. The bit rates for these networks are in most cases lower than the bit rates of wireline broadband access networks. As users become more and more accustomed to enjoying the low cost and high speed of wireline broadband access, the bit rate expectations of future wireless networks increase. One way of increasing the bit rates in future wireless networks is to use scheduling algorithms that exploit rapid signal strength fluctuations at the receiver. The purpose of this thesis is to evaluate existing scheduling algorithms and propose improvements such that throughput and resource utilization can be increased. Before the performance of the scheduling algorithms can be evaluated it is necessary to define how performance should be measured. This thesis, therefore, proposes how fairness should be defined for scheduling algorithms exploiting multiuser diversity. This thesis also proposes how fairness should be measured on both layer 2 and layer 4. Two new scheduling algorithms that exploit multiuser diversity are defined in this thesis: opportunistic proportional fair and the deadline scheduler. Opportunistic proportional fair is designed for best effort services. Using numerical evaluation, opportunistic proportional fair is compared with a standard scheduling algorithm from the literature called proportional fair. For the evaluated scenario, opportunistic proportional fair has slightly higher throughput and slightly better fairness than proportional fair. The deadline scheduler is designed to handle a mixture of traffic flows for real-time and best effort services. Using numerical evaluation, the deadline scheduling algorithm is compared with some of the most promising algorithms proposed in the literature, including proportional fair, exponential rule and CDF-based scheduling. The algorithms are evaluated in a number of different traffic load scenarios in a multiservice wireless network. In the evaluation, the deadline scheduler is the only algorithm that performs really well in all scenarios.