T.J. Kim
fair queuing, WFQ, packet scheduler, latency, timestamp
Fair queuing algorithms providing Quality-of-Service (QoS) guarantees can be classified into two design methods in terms of the reference time used in calculating the timestamp of arriving packet: Start-Time (ST) and Finish-Time (FT). It was known that the former, even though yielding better bandwidth utilization, does not satisfy the requirements of isolating flows and providing differentiated QoS guarantees, while the latter is still suffering from a poor utilization. This paper proposes a ST-based algorithm satisfying those requirements, called Latency-Differentiable Start time Fair Queuing (LDSFQ), in which the architecture of multiple ST servers in parallel is introduced. The evaluation results show that the LDSFQ scheduler yields better bandwidth utilization by up to 45% than that in the FT one under a practical internet phone scenario.
Important Links:
Go Back
