Analytic Performance Model of P Fault-Prone Parallel Processors Running N Deterministically Distributed Tasks

摘要

This paper focuses on the analytical performance analysis off fault-prone processors that are running a job consisting of N independent tasks in parallel. All the tasks are assumed to have the same running time. We use a discrete Markov model for the analysis of the system. Assuming that there is no arrival process for tasks, the performance parameters that we calculate are conditioned on the job finishing successfully. We also assume that all the processors can fail but if the number of tasks is less than the number of processors, then the idle processors cannot fail. We will derive expressions for the mean execution time and the probability of success of a job. The results will be presented as probability distributions and graphs. Finally, we will compare our results with a similar system of processors running tasks having exponentially distributed times, as reported in [1].