IP网络时延敏感型业务流自适应负载均衡算法

Towards load adaptive routing based on link critical degree for delay-sensitive traffic in IP networks

互联网对时延敏感的业务数据流,要求具有较低的端到端时延,但是网络拥塞的发生,将会使服务质量无法保证。基于链路关键度提出了一种新的自适应负载均衡路由算法(LARA,load adaptive routing algorithm),能最大限度地避开拥塞链路从而减少端到端延迟。该算法通过得到一个优化目标函数,并利用凸优化理论将优化目标函数分解为若干个子函数,最终得到一个简单的分布式协议。利用NS2仿真器在基于CERNET2真实的拓扑结构上进行仿真实验,同时与网络中能普遍部署的等开销多路径(ECMP,equal-costmulti-path)算法相比较,通过测试反馈时延、分组丢失率、流量负载,结果表明LARA具有更好的自适应性和健壮性,性能相比更优。

Delay-sensitive traffic requires lower end-to-end delay in IP networks, such as online video, VoIP, video con- ference. Based on the criticality degree of link. A load adaptive routing algorithm （LARA） was presented which could avoid the link to be congested to reduce the end-to-end delay. Firstly, an optimization objective function has been put forward; and then decomposed into several sub-functions by using convex optimization theory; finally, the optimization objective function and sub-functions were transformed into a simple distributed protocol. LARA with ECMP （equal-cost multipath） routing strategy was compared which was widely deployed in the network by using NS2 simulation under CERNET2 topology. By evaluating the feedback delay, packet loss rate and traffic load, the results show that LARA can exhibit good performance and achieve excellent load balance, and meanwhile improve the robustness of the link when using multipath routing technology.