Quality of service scheduling for heterogeneous data services on wireless systems

With the rapid growth of the Internet and other IP-related applications, wireless systems have increased their range of supported services from circuit-switched voice to a wide variety of packet-switched data services. The introduction of data services has significantly increased the strain on wireless systems due to increased load as well as stricter and more diverse QoS requirements. The increased load and heterogeneity of QoS requirements make providing good service extremely difficult.; This dissertation introduces a class of packet data scheduling algorithms which increases system throughput and provides improved QoS support for heterogeneous data services. Data services contain a variety of QoS parameters. The most common are traffic handling priority and maximum packet delay. However, user throughput, blocking, and dropping are also of interest. In an effort to satisfy these varied service requirements, the algorithms adapt to varying packet delays on the link layer and channel bitrates on the physical layer in order to meet packet delay deadlines while increasing throughput.; The Adaptive Cross-Layer scheduler is the first algorithm introduced in this class. It determines the scheduling order that minimizes the delay measure of packets in the queue based on a cost function that takes into consideration the varying packet lengths, delay requirements, and channel bitrates. This cost function evaluates the impact of varying packet channel bitrates on the delay measure and signals to the scheduler when taking advantage of high packet bitrate opportunities improves delay performance.; The Parameterized Adaptive Cross-Layer scheduler is another algorithm in this class. It introduces service class and channel quality sensitivity parameters into the cost function, which adjust the scheduling priority of different service classes and modulate the scheduler's sensitivity to varying packet channel bitrates. Adjusting the service class weights allows the scheduler to improve the performance of one service class at the expense of another, while the channel quality exponent adjusts the prioritization of high channel quality packets.; Results indicate that the Adaptive Cross-Layer and Parameterized Adaptive Cross-Layer scheduling algorithms significantly outperform a number of common data scheduling algorithms with respect to delay, throughput, blocking, and dropping despite access to only rough estimates of channel quality.