WEB SERVER WORKLOAD PREDICTION: FUZZY MARKOVIAN APPROACH

摘要

In this article Web server workload is modelled by considering the most significant metrics, the burstiness factor and the bandwidth utilization. The integration of the workload metrics is accomplished by using a fuzzy inference system (FIS) to lessen the complexity of the metrics relation. Due to the fluctuating of WWW requests, the utilization states are viewed as a series of stochastic processes. The fuzzy Markovian prediction is a study in two manners: first, the utilization states, and second, the Markovian property. The use of the fuzzy approach for numerical calculation is demonstrated on an example of a real web server LOG data. The results of fuzzy approaches constitue a more appropriate comparison with the traditional approach.