Ant Colony Algorithm for Scheduling Parallel Program Based on DAG Graph Heuristics

Decreasing the execution time of parallel program is a key issue in the effective utilization of multiprocessor systems, and it is also a NP-hard problem. In this paper, we propose different static and dynamic attributes of task graph as ant colony heuristics to deal with the multiprocessor scheduling problem. The attributes based on task graph DAG (direct acyclic graph) for parallel program, including static and dynamic, express precedence constraints as well as computation cost and data transferring cost between tasks when they are scheduled to different processors. The heuristics in artificial ant colony algorithm is an important factor to solve combinatorial optimization problems. Analyzing the rationale behind these heuristics, we combine the problem-space heuristics and the search ability of ant colony algorithm efficiently for scheduling problem. The results demonstrate some heuristics perform better than other heuristics and dynamic heuristics beyond static heuristics and some classical list scheduling algorithms