OpenFlow-based Server Cluster with Dynamic Load Balancing
Hsien-Yi Liu, Ching-Yi Chiang, Hui-Sheng Cheng, Mei‐Ling Chiang · 2018
It is always a critical issue for IT managers to balance the servers' load in a server cluster. Nowadays, Software-Defined Network (SDN) is a novel way for data centers to manage their network environments with better flexibility, reliability, and performance. Therefore, more and more studies have focused on Software-Defined Network and designed better load balancing strategies for modern network environments. In this study, to fully utilize the hardware resources, we propose Least-Loading (LL) load-balancing strategy based on OpenFlow which is one of the first and most widely used SDN standards. The proposed LL strategy distributes requests to servers according to the real-time status (i.e. CPU utilization, memory utilization, and disk utilization) of each server in the server pool. Moreover, by reducing the overhead of selecting request handling server, our LL strategy avoids making the controller a system bottleneck and effectively improves the entire system throughput. Experimental results show that our LL strategy is 17.2% and 21.4% better than the original load-balancing strategies (i.e. Round-Robin and Weighted Random Selection) in Floodlight controller.