On maximum rate control of worst-case weighted fair queueing

Maximum rate control in a shared channel is important to service providers and carriers for various reasons. Previous approaches either used a concatenation of regulator and scheduler, which employs two set of queues and two management systems, or a policer in front of scheduler. The former requires extra buffer space and high overhead, and the latter causes inaccuracy. In this paper, we propose a new scheduling algorithm, called WF²Q-M (worst-case fair weighted fair queueing with maximum rate control), to simultaneously support maximum rate control and provide minimum service rate guarantee. WF²Q-M employs a WF²Q like scheduler without policer or regulator that it is designed to provide accurate scheduling with low overhead. We prove that in WF²Q-M the packet´s eligible time can be merged into its virtual starting time, and propose virtual clock adjustment to distribute the excess bandwidth of saturated sessions to other sessions without recalculating their virtual starting and finishing times. We also prove that WF²Q-M performance is theoretically bounded by a fluid reference mode, and experiments show WF²Q-M performs just as claimed.