SDN中基于PSVD的交换机迁移算法

A Dynamic Switch Migration Algorithm based on PSVD in SDN

针对软件定义网络中控制器管理的交换机流量分布不均衡问题,提出一种基于奇异值分解性能预测(PSVD)的交换机动态迁移算法。首先结合当前负载以及过去相似时刻的负载来预测未来时刻的控制器负载信息,然后将目前具有最大资源占用率的交换机迁移至低负载且未来时刻仍然处于低负载状态下的目标控制器。实验结果表明,该算法提高了控制器的资源利用率,实现了控制器的负载均衡,同时缩短了控制器的响应时间。

Aiming at the unbalanced flow of the switch managed by the controller in the Software Defined Network （SDN）, a switch migration selection algorithm based on Performance Prediction of Singular Value Decomposition （PSVD） is proposed. By combining the current load and the load at a similar time in the past to predict the load information of the controller in the future, the switch with the maximum resource occupancy is then moved to a target controller that is currently under light load and still under a low load in the future. The experimental results show that this algorithm improves the resource utilization of the controller, realizes load balancing and shortens the response time of the controller.