In this paper we consider the problem of scheduling a given task graph on a dynamic network, where processors may become available or unavailable during the lifetime of the computation. We show that known list scheduling algorithms which use task cloning can be extended to develop efficient algorithms in this model. We als present a different approach where in anticipation of processor failures and recoveries, a set of schedules are precomputed and schedule switching is done whenever a failure or recovery takes place.
展开▼
