FAST: a low-complexity algorithm for efficient scheduling of DAGs on parallel processors

The DAG scheduling problem is a rich land of research and a plethora of algorithms for solving this problem have been reported in the literature. However, designing a scheduling algorithm of low complexity without sacrificing performance remains a challenging obstacle from a practical perspective. In this paper, we present a local search-based scheduling algorithm that attempts to meet this challenge. The proposed algorithm is called Fast Assignment using Search Technique (FAST). Its overall time complexity is only O(e) where e is the number of edges in the DAG. The algorithm works by first generating an initial solution and then refining it using local neighborhood search. The algorithm outperforms numerous previous algorithms while taking dramatically smaller execution times. The distinctive feature of our research is that the performance evaluation is not carried out using simulation, rather we have tested our proposed algorithm and compared it with other algorithms using a parallel compiler with real applications on the Intel Paragon.