Efficient Process Scheduling for Multi-core Systems

Machines with multiple cores have become more and more popular. In order to fully utilize their parallel computation ability, efficient scheduling algorithm plays an important role. A good scheduler should output the reasonably good result quickly but most of the current schedulers fail to achieve this goal and always have to compromise between the running time and result quality. In response to the above concerns, this paper proposes one algorithm, Longest Path First In (LPFI), to do scheduling efficiently and effectively for multi-core. This algorithm uses a deterministic allocation mechanism to prioritize processes which are in long dependency chain. The experiment results show that, compared with greedy scheduling algorithm, LPFI has around 10% improvement in the final result and can output the optimal result much faster than integer linear programming (ILP) scheduler.