采用子网流量组合优化的网络流量分割方法

A Traffic Splitting Method of High-Speed Network Using Sub-Network Flow Optimization

针对当前基于仿真的流量分割方法存在处理开销过大且仅能保证流层面流量完整性的问题,提出了一种组合优化的网络流量分割方法.将流量经分割后得到的数据包总延迟次数作为优化目标,由此网络流量分割问题转化为二次分配优化问题;用子网流量作为基本单元进行分配,可在减小问题求解规模的同时保证子网流量的完整性.利用真实网络流量对所提方法的分割效果进行了验证,并与基于数据包和流的轮询流量分割方法进行了比较,结果表明,所提方法能够达到基于流的轮询流量分割方法的性能,数据包延迟较之不采用优化的基于子网流量轮询方法显著减小,性能提高近10%.

The aim of traffic splitting in parallel traffic replay is to distribute high-speed network traffic to multiple replay nodes such that the discrepancies between the generated traffic and the original traffic can be minimized.Currently adopted simulation based methods have a high computational cost and can only guarantee flow level traffic integrity.In this paper,a new method for traffic splitting of high-speed network is proposed based on packet delay time minimization.The problem is converted into a quadratic assignment problem.The computational cost can be dramatically reduced and sub-network level traffic integrity can be achieved by aggregating flows from the same sub-network and assigning traffic in unit of sub-network flows.The method is validated by using actual traffic traces,and a comprehensive comparison between the proposed method and other available traffic splitting methods such as packet based and flow based round-robin approaches is carried out.Empirical studies show that the proposed method can achieve the same replay performance as the flow-based round-robin approach,and achieves 10% performance enhancement over the other approaches with the same split unit.