Supporting Best-Effort Traffic with Fair Service Curve

摘要

Packet Fair Queueing (PFQ) algorithms are the most popular and well studied scheduling algorithms for integrated services networks for two reasons: (1) With reservation, they can provide per-flow end-to-end delay guarantees for real-time traffic flows. (2) Without the reservation, they can provide protection among competing best-effort flows while allowing dynamic bandwidth sharing. However, PFQ algorithms have two important limitations. The first one is that, since only one parameter (a weight) is used to allocate resource for each flow, there is a coupling between delay and bandwidth allocation. This can result in network under-utilization when real-time flows have diverse delay and bandwidth requirements. The second and less well known limitation is that, due to the instantaneous fairness property of PFQ algorithms, when used for best- effort service, PFQ algorithms favor continuously- backlogged throughput- oriented applications such as FTP over bursty applications such as WWW and telnet. In a previous study, we proposed the Fair Service Curve (FSC) algorithm which enables more flexible delay and bandwidth allocation for real-time traffic through the use of non-linear service curves. In this paper, we show that, when used for best-effort traffic, FSC can improve performance of delay-sensitive bursty applications without negatively affecting the performance of throughput-oriented applications.