Sensitivity of website reliability to usage profile changes

Abstract

To measure the reliability of a website from a user's point of view, the uncertainty on the usage of the website has to be taken into account. In this paper we investigate the influence of this uncertainty on the reliability estimate for a web server. For this purpose a session based Markov model is used to model the usage extracted from the server's logfiles. From these logfiles a complete user profile can be extracted together with an estimate of the uncertainty on this user profile. This paper investigates the applicability of this kind of Markov model on web server reliability and discusses the difficulties with data extraction from the logfiles. Advantages and disadvantages of this approach are discussed and the approach is applied to data from a university department's web server to demonstrate its applicability. © 2007 IEEE.