An Algorithm and Some Numerical Experiments for the Scheduling of Tasks with Fault-tolerancy Constraints on Heterogeneous Systems

In this paper, we propose an efficient scheduling algorithm for problems in which tasks with precedence constraints and communication delays have to be scheduled on an heterogeneous distributed system with an one fault hypothesis. Based on an extension of the Critical-Path Method CPM/PERT, our algorithm combines an optimal schedule with some additional tasks duplication, to provide fault-tolerance. Backup copies are not established for tasks that have already more than one original copy. The result is a schedule in polynomial time that is optimal when there is no failure, and is a good resilient schedule in the case of one server failure. We finally compare the optimal solutions with the resilient solutions found by this algorithm on several semi-random DAGs.