Multi-stage scheduling scheme for massively parallel systems

Multi-stage scheduling has been proposed as a hierarchical scheduling technique to achieve load balancing in massively parallel multiprocessor networks. Existing hierarchical approaches can not easily map on multiprocessor system because they do not incorporate the inherent parallelism of a multiprocessor system when task allocation is made. In this paper a hierarchical based novel scheduling scheme named as Multi-stage scheduling has been proposed and implemented on multiprocessor systems. The scheme divides the whole network into various segments, where each segment consists of a set of processors and makes the load balancing in an incremental way. The performance of the proposed scheme is evaluated and a comparison with the existing scheme is made. The load imbalance on a particular multiprocessor system is calculated in terms of Load Imbalance Factor (LIF). Simulation results shows that the proposed multi-stage scheduling gives better performance in terms of task scheduling on different multiprocessor interconnection networks.