Performability Analysis of a Redundant Parallel Task in Network Systems

Recently, application of network systems ( e.g. , cloud computing systems) is increasingly prevalent for achieving integration, sharing and efficient utilization various resources. A parallel task in a system has multiple subtasks that can be executed different servers parallel. However, failures any subtask inevitably result the entire cannot complete. To avoid such situation, create some copies from make them run on simultaneously. This redundant execution manner an approach to improve performance guarantee reliability. it also brings complexity modeling, evaluation optimization. For example, link lead inaccessibility servers, server hosted complicated failure correlation ignored modeling evaluation. paper first presents reliability-performance model system. The captures precedence constraints subtasks, types correlations fidelity. design algorithm encompasses Graph theory Bayesian theorem evaluate performability metric, which used quantify important correlation. Finally, heuristic designed search optimal strategy maximizes metric. Illustrative examples are presented.