高速网络性能测试负载生成多流发送调度算法  

Flow scheduling algorithm for traffic generation in high-speed network performance testing

作　　者：盛立杰[1] 吴建平[1] 章淼[1] 徐明伟[1] 

机构地区：[1]清华大学计算机科学与技术系,北京100084

出　　处：《清华大学学报（自然科学版）》2004年第7期969-973,共5页Journal of Tsinghua University(Science and Technology)

基　　金：国家"八六三"高技术项目(863-300-05-04-01-00);国家自然科学基金资助项目(90104002;60303006;60203025);教育部科学技术研究重点项目(02004)

摘　　要：测试负载生成是网络性能测试的重要环节之一。为了在测试过程中模拟复杂的数据流量环境,常常通过组合大量特征各异的流来近似实际网络流量。流的重要属性是其速率与突发程度,文中将此定义为流模式。在使用多个流来模拟复杂的网络环境时,一个重要问题是如何调度多个流,以保证各个流的流模式。提出了一种调度方法GRR(groupingroundrobin),将所有流分成若干小组后分别在组内和组间两级轮询调度,可在较低的时间和空间复杂度内实现调度目标。算法已在一个高速网络性能测试系统IP-TEST中得到应用,达到了1Gb/s的线速发送速率,并且支持突发流模式。Traffic generation is a key component of high-speed network performance testing. A complex data traffic environment can be emulated for performance testing by generating simultaneously a number of flows with various data characteristics. The data rate and the degree of bursting of the packets in each flow are important parameters that are specified by the flow model in this paper. Since many flows should be sent from each network interface, a scheduling mechanism is needed to maintain the flow model of each flow. This paper describes a scheduling algorithm that groups the flows and polls them in round robin in both the flow-level and the group-level. The algorithm can achieve the scheduling objective with low time and space complexity. The algorithm implemented in a high-speed network performance tester, IP-TEST achieved a line-speed of 1 Gb/s and supported burst flow models.

关 键 词：计算机网络 网络性能测试 测试负载生成 流调度 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]