一种集群系统的透明动态反馈负载均衡算法

随着服务器集群系统大量应用于各中小企业的信息系统中，传统均衡算法一方面由于局限性达不到企业的要求，另一方面大部分中小型企业也无法承受昂贵的硬件负载均衡器费用，本文提出了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，同时引入一个负载冗余以动态调整节点负载分配，从而达到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和提高系统响应时间的目标。测试表明，该算法有效的提高了系统服务性能，且优于静态分配算法和Pick-KX算法。

A Transparency Dynamic Feed-back Load Balancing Algorithm Based on Cluster System

With the broadly adoption of server cluster systems in the enterprise information system, the traditional algorithm of load balancing can’t satisfy the requirement of enterprise use because of either the expensive hardware cost or the limitation of algorithm itself. In this paper, a cluster based transparency dynamic feed-back load balancing algorithm (TDLBA) is presented. To simplify the distribution of the load balancer, improve the throughput rate and response time of the system, this algorithm not only considers multi-resources abundantly (such as CPU、memory、I/O and network bandwidths etc.), uses redundant backup strategy, adopts the dynamic feed back algorithm of the server nodes, but also brings in a redundant load node for the adjustment of load allocation. Conclusions can be drawn through test that this algorithm improves the service capability of the system effectively compared with static allocation algorithm.