关键词: 分组交换机, 负载均衡, 调度, 多播, 两级交换结构

Abstract:

This study focuses on the scalability problems for very large, high-speed switches, and presents a two-stage integrated scheduling algorithm named UMCSA(Unicast and Multicast Concurrent Scheduling Algorithm)which supports both unicast and multicast traffic simultaneously. The first stage of the switching fabric performs switching for unicast traffic and load-balancing for multicast traffic, while the second stage performs switching for load-balanced multicast traffic. With a two-phase(request-grant)scheme, the proposed algorithm performs without iteration, and at the same time reduces the scheduling overhead to O(logN) significantly. By using the VOQ(Virtual Output Queuing)technique for unicast and multicast traffic separately at different stages, the HoL(Head-of-Line)blocking is eliminated and therefore a good throughput performance could be achieved. UMCSA performs scheduling for unicast and multicast traffic in parallel at different stages with simple Round-Robin arbitration, which is more suitable for high-speed applications. Simulation results show that the proposed integrated algorithm exhibits a good performance in terms of throughput and average delay, at different traffic compositions under various traffic patterns.

Key words: packet switches, load-balancing, scheduling, multicast, two-stage switch architectures