Introspective failure analysis: avoiding correlated failures in peer-to-peer systems
Hakim Weatherspoon, T. Moscovitz, John D. Kubiatowicz · 2003
Failure independence is an important assumption for many fault tolerance techniques. Unfortunately, real systems exhibit correlated failures. In this paper, we present a framework for online discovery of groups of server nodes that are maximally independent in their failure characteristics. We discuss the framework in detail and provide a preliminary evaluation.