An island-based evolutionary algorithm for maximizing schedule reliability

The paper investigates the problem of scheduling multiple-variant (m-v) tasks on multiple processors with a view to maximize schedule reliability. Schedule reliability is understood as the probability that all tasks within schedule are executed without failures. It is proposed to manage complexity issues of the m-v tasks not at a single task level, but from the point of view of the whole set of tasks to be run under some time constraints. Because the problem is NP-hard, two evolutionary algorithms - an island based and a traditional population based algorithm to solve the problem have been proposed. The approach seems promising and the algorithms produce good to satisfactory solutions.