Authors: Xiao Xiao; Tadashi Dohi
Addresses: Department of Information Engineering, Graduate School of Engineering, Hiroshima University, 1-4-1, Kagamiyama, Higashi, Hiroshima, 739-8527, Japan ' Department of Information Engineering, Graduate School of Engineering, Hiroshima University, 1-4-1, Kagamiyama, Higashi, Hiroshima, 739-8527, Japan
Abstract: In this paper we focus on the relationship between the error rate which is one of the representative reliability measures in Apache web servers and the system parameters which reflect on the web server's system performance, and develop regression-based probability models to describe it. More specifically, we implement a simple client server system and carry out an experiment to measure both the error rate and the system parameters. As the result of quantitative evaluation on the proposed models, it is shown that our models could fit the empirical error rate even when the number of threads varies. In addition, we select out seven most effective system parameters based on a backward selection and prove that they can contribute much more to the error rate rather than the other combinations of system parameters in terms of prediction assessment.
Keywords: web server reliability; Apache web server; reliability measurement; dependability evaluation; error rate estimation; system parameters; regression modelling; backward selection; prediction assessment; probability modelling; web servers.
International Journal of Reliability and Safety, 2013 Vol.7 No.1, pp.1 - 16
Accepted: 28 Apr 2011
Published online: 06 Aug 2013 *