Search

Web application modeling offers several challenges; handling a composition of several languages, both server and client side code; handling state dependent dynamism; and handling immense user interaction possibilities. In this paper we explore the possibilities of developing a Markov model of a part of a web application in order to get an estimation of the system's reliability. The model is developed from a control flow graph of code components. The transition probabilities are obtained from estimating the probability of a user's status and from the internal structure of the code. Input testing results are used to estimate the reliability of specific states. Also, a weight is used in the reliability estimations, stemming from modeling a state visited at different occasions as one state. The type of reliability modeled concerns source failure. Failures related to browser, host or network failures are not considered. The framework developed is a general and straight-on way to get a quantitative measure of reliability. The developed concepts can be used on an application with similar features i. e. storing user submitted information. The advantages of the model are dynamic, depending on the particular features of exactly that implementation and on what part of the system that is modeled. [ABSTRACT FROM AUTHOR]