软件定义网络多控制器部署优化方法

Optimal Placement of Multiple Controllers in Software Defined Network

目前软件定义网络(software defined network,SDN)多控制器管理的研究重点和难点在于多控制器部署,合适的控制器数目和合理的部署位置能够实现低延迟和高可靠性的折中。在网络拓扑图上将部署位置问题简化并形式化,定义了可靠性和延迟指标,设计了多控制器部署位置求解算法(multi-controller placement location algorithm,MCP)。MCP通过遗传算法将网络划分为多个子网,并在每个子网中放置一个控制器,以最大化网络可靠性和最小化路径延迟为目标,通过模拟退火算法确定控制器在每个子网中的位置,经对比实验验证了MCP生成的部署位置方案在可靠性和延迟上的性能优势。以可靠性和延迟为评估指标,基于MCP测试了各种网络规模的最合适控制器数目,并分析得出最合适控制器数目与网络规模之间的统计规律。

The current research focus and difficulty of multi-controller management in software defined network(SDN)is the multi-controller placement.Suitable number of controllers and reasonable placement of controllers can reach a compromise between low latency and high reliability.The problem of controller placement was simplified and visualized on the network topology chart,the indicators of reliability and delay were defined,and the multi-controller placement location algorithm(referred to as MCP)was designed.With the genetic algorithm,MCP divided the network into some subnets,and every subnet was placed with a controller.With the objective to maximize network reliability and minimize path delay,the simulated annealing algorithm was used to determine the location of the controller in every subnet.The performance advantages of MCP in terms of reliability and delay were verified by comparative experiments.With reliability and delay as evaluation indicators,a lot of experiments based on MCP have been conducted and the statistical law between the most suitable number of controllers and the network size has been obtained.