Performance degradation or system resource exhaustion can be attributed to inadequate computing resources as a result of software aging. In the real world, the workload of a web server varies with time, which will cause a nonlinear aging phenomenon. The nonlinear property often makes analysis and modelling difficult. Workload is one of the important factors influencing the speed of aging. This paper quantitatively analyzes the workload-aging relation and proposes a framework for aging control

more »

... der varying workloads. In addition, this paper proposes an approach that employs prior information of workloads to accurately forecast incoming system exhaustion. The workload data are used as a threshold to divide the system resource usage data into multiple sections, while in each section the workload data can be treated as a constant. Each section is described by an individual autoregression (AR) model. Compared with other AR models, the proposed approach can forecast the aging process with a higher accuracy.