BCSQ: Bin-based Core Stateless Queueing for Scalable Support of Guaranteed Services

摘要

Core stateless packet scheduling systems have received considerable attention in recent years because of their scalability in supporting per-flow Quality of Services guarantees. In such a system core routers do not need to maintain per-flow state and do not need to perform per-flow operations such as per-flow classification, per-flow queueing, and per-flow scheduling. On the other hand, existing core stateless packet schedulers require core routers to sort incoming packets based on their virtual finish times. This sorting operation results in the worst-case runtime complexity of O(log_2N), where N is the number of packets in a scheduler. In this paper we propose a bin-based core stateless queueing (BCSQ) algorithm, which achieves constant runtime complexity that is independent of the number of packets in the scheduler. We present the detailed design of BCSQ and derive the worst-case end-to-end delay bounds for packets in a network of BCSQ. In addition, we investigate the effects of the configurable parameters of BCSQ on the performance of BCSQ networks. Simulation studies are also performed to illustrate the efficacy and performance of BCSQ.