Survival of the Internet applications: a cluster recovery model

Abstract

Internet applications become increasingly widely used for millions of people in the world and on the other hand the accidents or disruptions of service are also dramatically increasing. Accidents or disruptions occur either because of disasters or because of malicious attacks. The disasters could not be completely prevented. Prevention is a necessary but not a sufficient component of disaster. In this case, we have to prepare thoroughly for reducing the recovery time and get the users back to work faster. In this paper, we present a cluster recovery model to increase the survivability level of Internet applications. We construct a state transition model to describe the behaviors of cluster systems. By mapping through recovery actions to this transition model with stochastic process, we capture system behaviors as well as we get mathematical steady-state solutions of that chain. We first carry out for steady-state behaviors leading to measures like steady-state availability. By transforming this model with the system states we compute a system measure, the mean time to repair (MTTR) and also compute probabilities of cluster systems failures due in face of disruptions. Our model with the recovery actions have several benefits, which include reducing the time to get the users back to work and making recovery performance insensitive to the selection of a failure treatment parameter