Enterprise web site problem diagnosis using Bayesian belief networks

Enterprise web sites are characterized by large volumes of traffic, multi-tier architectures, multiple COTS software products, and distributed system design. Diagnosing problems in such environments requires expertise gained through extensive experience. Experienced subject matter experts are expensive. This paper describes a decision support system based on Bayesian belief networks to diagnose the problems of a typical enterprise web site that utilizes HTTP servers, J2EE compliant application servers, database servers, firewalls, messaging servers, and single sign-on.