A fuzzy algorithm for adaptive multilevel queue management with QoS feedback

Multilevel and multiclass queuing is very common in systems having QoS related concerns and one of the most common and well-know approaches is Weighted Fair Queuing (WFQ). In WFQ there are multiple queues and each one is dedicated for a distinct class of jobs or traffic and there is a weight value assigned to each queue to indicate the number of jobs that has to be picked from the queue in next round. This approach uses fixed predefined weight values tuned by system administrator with knowledge about system's characteristics. In this paper we extend WFQ approach by adding a fuzzy inference system and QoS measurement component. The proposed algorithm uses its easy-to-extend and agile knowledge base to configure the queue weights dynamically and autonomously in reaction to recent QoS metrics to achieve better performance and improved QoS. The Proposed method tailored to be used in virtual resource management environments especially in Grid services. To evaluate the performance of the proposed algorithm, simulation software used to compare it with well-known queue management approaches. The simulation results confirm the validity of the proposed algorithm.