New heuristics and integer programming formulations for scheduling divisible load tasks

Divisible load applications occur in many fields of science and engineering. They can be parallelized in a master-worker fashion, but they pose several scheduling challenges. We propose single-round and multi-round integer programming formulations for scheduling divisible load applications with minimum makespan. An innovative linear-time exact algorithm improving the complexity of the best known algorithm to date is described for the special case in which the processor activation order is known beforehand. This algorithm is embedded within a greedy-with-feedback heuristic for finding good solutions for single-round problems. Numerical results illustrate the speed and the effectiveness of the proposed heuristic.