| Summary: | 碩士 === 國立成功大學 === 工程科學系 === 88 === ABSTRACT
By advances in network architecture and multimedia applications, people have more and more desire on the quality of service in Internet. Because that current Internet provides same-service-to-all model and the difference between users/applications requests is much large, the congestion and unfair resource allocation make users not satisfy with the quality of service. Thus many studies focus on this problem and hope to provide sound solutions. There into, the differentiation service model is one of possible ways to implement. The main idea of this model is to aggregate these flows of similar quality requests into a class and network elements (i.e., routers, switches) only maintain per-class QoS and resource allocation. Because of the different quality levels under each class, users/applications could select the most suitable class based on his own request or class pricing.
In this thesis, we provide the architecture of the proportional differentiation service, and hope to solve the problem by its controllability (network operators can adjust the quality spacing between classes) and predictability (anytime higher classes get better service than lower classes). About the forwarding mechanism inter routers, we also provide the dynamic weighted fair queuing scheduler and take the short-term queuing delay as our measure QoS parameter. By using the method, the proportional differentiation becomes more service feasible. Furthermore, simulation results show that the proposed methods can realize proportional delay differentiation service effectively.
|
|---|
