Jitter Controlled Quality of Service with Measurement Based Admission Control

摘要

Providing reasonable quality of service (QoS) for real-time applications without making absolute guarantee is studied extensively due to its higher resource utilization than guaranteed service. Improving network resource utilization strongly relies on a traffic scheduling method that is adaptable to varying traffic load and an admission control algorithm that does not depend on worst-case traffic description. This paper presents a jitter controlled scheduling approach called LAZY with a measurement-based admission control algorithm to study the effect of different holding times under different traffic loads on end-to-end delay and jitter. LAZY approach provides on-time delivery service. Hence, packets arrive at the destination neither too late nor too early, but right around the deadline. This service could ensure an end user that he would receive the requested data when he needs them. In addition, network bandwidth could be utilized more efficiently. Our simulation results shows that under light and moderate traffic load the LAZY method could maintain comparable effectiveness of the bandwidth yet achieve better time delivery relative to the target deadline compared to ones of the conventional FIFO strategy. But under heavy traffic load, the LAZY method accepts more traffic flows yet still attains time delivery closer to the target deadline than the FIFO approach does.