Evaluation of eligible jobs maximization algorithm for DAG scheduling in grids

Among many attempts to design DAG scheduling algorithms that would face grid environment requirements, the strategy of number of eligible jobs maximization seems promising. Therefore, this paper presents the results of thorough analysis and evaluation of this strategy and its implementation called PRIO. We have analysed a large space of random DAGs and various resources parameters to compare results of PRIO algorithm with standard critical path length prioritization, FIFO prioritization as well as with quasi-optimal solution. Results of this comparison, in terms of the makespan and robustness, are supplemented by a theoretical and specific case analysis. We conclude with an assessment of usefulness of the current implementation of eligible jobs maximization strategy.