混合SDN的多目标路由优化算法

Multi-Object Routing Optimization Algorithm for Hybrid SDN

为充分利用SDN(Software Defined Network,软件定义网络)节点的优势来优化网络性能,混合SDN的流量工程成为当前的研究热点,而路由优化是实现流量工程目标的重要策略之一。但是,当前混合SDN的流量工程中未考虑网络整体的负载均衡以及SDN节点的处理能力。针对上述问题,提出了一种多目标路由优化算法——MCS(Minimum Cost Sum,最小化代价和),综合考虑整个网络的传输延迟与链路利用率,同时保证SDN节点的处理能力满足实际约束,最终实现全网综合性能的最优化。实验结果表明,当网络整体负载较轻,MCS与现有的SOTE(SDN/OSPF Traffic Engineering,软件定义网络/开放最短路径优先流量工程)算法性能相近;而当网络负载加重时,MCS相比于SOTE,可将网络负载降低约9%,因此,MCS算法具有更高的优化能力。

To make full use of advantages of Software Defined Network（SDN）nodes to optimize network performance,traffic engineering of hybrid SDN becomes a focus of current research,and routing optimization is one of the key strategies to realize the goal of traffic engineering.However,current traffic engineering of hybrid SDN fails to take into consideration the overall load balancing of network and processing capacity of SDN nodes.To solve the problem,a multi-object routing optimization algorithm,known as Minimum Cost Sum（MCS）was proposed to optimize link utilization and transmission delay of SONet under the actual restriction of SDN-FE.Experiments show that in light network load,MCS is similar to SDN/OSPF Traffic Engineering（SOTE）in performance,but the former could reduce 9% of network compared with the latter in heavy load.Therefore,MCS is able to provide better optimization.