Availability modeling and evaluation of web-based services

Overview & Synopsis

This thesis presents a pragmatic modeling approach allowing designers of web-based applications and systems to evaluate the service availability provided to the users. Multiple sources of service unavailability are taken into account, in particular i) hardware and software failures affecting the servers, and ii) performance degradation (overload of servers, very long response time, etc.). An hierarchical multi-level approach is proposed based on performability modeling, combining Markov chains and queueing models. The main concepts and the feasibility of this approach are illustrated using a web-based travel agency. Various analytical models and sensitivity studies are presented considering different assumptions with respect to the architectures, recovery strategies, faults, users profile and traffic characteristics.