Improving quality of service for next generation networks with active buffer management.

摘要

Improving quality of service (QoS) of Internet has been being widely recognized as one of the critical issues for the next generation networks. Internet Engineering Task Force (IETF) recommends active buffer management for next generation networks to improve the quality of service. The goals of active buffer management are improving throughput, eliminating global synchronization, providing low delay service and ensuring fair bandwidth sharing. This thesis studies the problems in improving quality of service for next generation networks with active buffer management.; Random Early Detection (RED) is currently recommended active buffer management scheme by IETF. However, studies have shown that RED has problems of low throughput, large delay/jitter, poor time response, unfairness, and inducing instability of networks etc. Researches have also shown that the performance of RED depends on its configuration parameters.; This thesis proposes and develops new scheme, models, and algorithm to improve the throughput, delay, jitter, and time response performance of active buffer management, as well as to eliminate global synchronization phenomena. In details, an adaptive active buffer management scheme, called. Double Slope Random Early Detection (DSRED), is proposed to improve throughput and delay performance of active buffer management both in homogeneous and heterogeneous network environment. Models that based on self-similar TCP traffic characteristics to determine optimal parameters for active buffer management to eliminate global synchronization and to provide low delay service are developed. A new algorithm, called Low Pass Filter/Over Drop Avoidance (LPF/ODA), is proposed to improve time response of active buffer management.; The results show that the proposed schemes, algorithms, and models significantly improve the performance of active buffer management and therefore TCP performance. Researches in this thesis provide valuable insight into studying, developing, implementing, and deploying active buffer management schemes in next generation networks.