一种集群系统的透明动态反馈负载均衡算法

Transparency dynamic feed-back load balancing algorithm based on cluster system

提出一种集群系统的透明动态反馈负载均衡算法(TDLBA)。该算法充分考虑集群系统中多种资源(CPU、内存、I/O和网络带宽等),采用双机热备份负载均衡器,服务器节点周期动态反馈方法,同时引入一个负载容余以动态调整节点负载分配,从而尽量简化负载均衡器的任务分配算法,最大限度满足系统最大吞吐率,减少系统响应时间。测试表明,该算法有效地提高了系统服务性能,优于静态分配算法和Pick-KX算法。

With the broadly adoption of server cluster systems in the enterprise information system, the traditional algorithm of load balancing can not satisfy the requirement of enterprise because of either the expensive hardware cost or the limitation of algorithm itself. A cluster-based transparency dynamic feed-hack load balancing algorithm （TDLBA） was presented. To simplify the distribution of the load halancer, increase the throughput rate and decrease the response time of the system, various resources were considered abundantly （such as CPU, memory, I/O and network bandwidth）, redundant backup strategy was used, the dynamic feed hack algorithm of the server nodes was adopted, and a redundant load node was introduced for the dynamic adjustment of load allocation. The test shows that this algorithm improves the service capability of the system more effectively compared with static allocation algorithm and Pick-KX algorithm.