Dynamic Load-Balance Strategy Based on Software Aging in Web Server Cluster System

Many studies reported that servers become aging after running a long time and the efficiency of server cluster system will degrade. This paper mainly focuses on modeling and analysis of dynamic load balance strategy adding software aging parameter into dynamic load balance strategy. Adding software aging into load balance strategy is proposed for the first time in this paper. A software aging algorithm based on robust locally weighted regression (Loess) is deployed to estimate aging trend and the threshold point used to compute software aging parameter. Experiments verify the effectiveness of the algorithm presented in this paper. The conclusions drawn from this paper will greatly benefit the application of software rejuvenation technique into load balance strategy, that is, it makes it easy to determine when to perform software rejuvenation in a server before it reaches the downtime. The results for the cluster system show that performance of the cluster can be effectively improved.