论文部分内容阅读
并行交换是新兴的交换技术,基于该技术能够利用小型交换模块来构建大容量的交换系统,例如太比特或更高容量的交换机.把带输入队列的并行交换称为带缓存并行交换(buffered parallel switch,简称BPS),重点研究其中并行且独立工作的交换模块之间的负载平衡问题.从不同角度出发,提出两种负载平衡的定义.基于两种定义,分别分析了BPS负载平衡的条件并提出分布式调度算法族.最后,提出一种简单而有效的调度算法,该算法能在无加速比BPS中同时满足两种定义,仿真实验结果表明了该算法的有效性和良好性能.另外,就算法的工程实现进行了讨论.
Parallel switching is an emerging switching technology that enables the use of small switching modules to build high-capacity switching systems such as terabit or higher-capacity switches. Parallel switching with input queuing is called buffered parallel switch, referred to as BPS for short), and focuses on the problem of load balancing between switching modules that work in parallel and independently. Two different definitions of load balancing are proposed. Based on the two definitions, the conditions of BPS load balancing are analyzed separately This paper proposes a family of distributed scheduling algorithms.Finally, a simple and effective scheduling algorithm is proposed, which can meet two definitions at the same time in non-speedup BPS.The simulation results show the effectiveness and good performance of the algorithm.In addition, The algorithmic engineering is discussed.