Generating bursty web traffic for a B2C web server

Abstract

The paper deals with the problem of emulating highly bursty Web traffic that can be observed at inputs of Web servers hosting online Web stores. This problem is related to the broad issue of Web server performance prediction and evaluation through simulation experiments. Based on up-to-date results on real Web server workload analyses a workload model has been proposed. It combines a model of a user session at a Business-to-Consumer (B2C) Web site with HTTP-level workload models for business and non-business Web servers. The proposed model has been implemented in a workload generator. Based on statistics registered during a simulation experiment, a burstiness factor has been computed for the generated workload, which has proven to be highly variable and bursty. © 2011 Springer-Verlag.