分布式软件定义网络中多域流量工程的路由优化方法

针对分布式软件定义网络（software-defined networking,SDN）中流量管理调度不均衡的流量工程问题,提出一种基于负载均衡的多控制域流量路由优化的解决方案。首先分析控制消息流量的组成、域内通信及域间通信规则;然后基于4种控制消息定义控制链路流量的构成,明确链路承载流量分为控制消息流量和业务流量,建立平衡控制器负载和最小化最大链路利用率的优化模型;最后基于域内通信和域间通信提出两层路由算法。为提高模型求解精度,进一步提出改进离散萤火虫算法求解最优路由。结合ABILENE网络和GEANT网络,分析控制消息流量、控制器负载和链路负载等评价指标。实验结果表明,优化模型能有效实现控制器和链路负载均衡,控制消息流量是流量工程重要组成部分。相比集中控制模式,扁平分布式控制模式的平均控制器负载降低47.3%,最大链路利用率相差不超过15%。

Routing optimization method for multi-domain traffic engineering in distributed software-defined networking

Addressing the challenge of unbalanced traffic management and scheduling in distributed software-defined networking (SDN) for traffic engineering, we propose a solution for traffic routing optimization across multi-control domains based on load balancing. Firstly, we define the composition of message traffic and the rules for intra-domain and inter-domain communication. Then, we establish an optimization model aiming at balancing controller loads and minimizing maximum link utilization. The model is based on the composition of control link traffic using four control messages, with link traffic divided into control message traffic and network traffic. Finally, we propose a two-layer routing algorithm based on communication rules. To improve the accuracy of the model solution, we introduce an improved discrete firefly algorithm. Evaluating the model using the ABILENE network and GEANT network, we assess indicators such as control message traffic, controller load, and link load. Experimental results show that the optimization model effectively balances loads between controllers and links, emphasizing the significance of managing message traffic in traffic engineering. Compared to centralized control modes, the average controller load in the flat distributed control mode is reduced by 47.3%, with the the maximum link utilization difference not exceeding 15%.