On guaranteeing the quality of service of conformant traffic in excess bandwidth allocation for shared access networks

The current implementation and operation of traffic control schemes by Internet service providers (ISPs) prevent subscribers from getting the benefits of full sharing of resources available in shared access networks, where the capability of allocating available bandwidth by a scheduler (e.g., weighted fair queueing (WFQ)) in an access switch is limited to traffic already shaped by token bucket filters (TBFs) per subscribers' service contracts; with this practice of ISP traffic control in shared access, we cannot expect any fundamental sharing of resources in the long term except for that in the short term controlled by the size of a token bucket. To address it, we recently proposed traffic control schemes based on core-stateless fair queueing (CSFQ) and token bucket meters (TBMs), which can allocate excess bandwidth among active subscribers proportional to their token generation rates. Using a common first-in, first-out (FIFO) queue for all packets, however, degrades the short-term performance of conformant traffic due to the presence of non-conformant packets already in the queue. Also, the rate estimation based on exponential averaging makes it difficult to react to rapid changes in traffic conditions. In this paper we study on how to guarantee the quality of service of conformant traffic in all time scales in excess bandwidth allocation for shared access. Specifically, we propose a new traffic control scheme based on deficit round-robin (DRR) and TBMs, whose advantages over the CSFQ-based one are demonstrated through simulation results.