A performance model for P2P computing systems

In this paper we develop a framework that can be used to assess the performance of (peer-to-peer) P2P computing systems whose applications require large amounts of CPU cycles. This is realized by constructing an analytic queueing model representative of these types of P2P systems and consists of P processing nodes that comprise a peer group. The queueing model is based upon a Markov chain with transitions representing task service times at the processing nodes and rates for nodes cycling between available and unavailable states. We assume that service times for tasks at the processing nodes in the P2P system are exponentially distributed. Using this model we: show how performance measures for these types of P2P applications can be calculated for any number of tasks and peers such as the expected work performed by P2P application, the expected number of peers available for requesting or processing tasks, and the expected time to complete a P2P application given a finite number tasks. In addition, we provide some numerical examples.