A Study On Parallel Task Allocating And Scheduling Technology Based On Message-Oriented Middleware

As the using of computer technology advances, real time and parallel computing are becoming the main stream of computation technology. At the same time, task allocating and scheduling algorithm is major factor of the performance of parallel computing system. It affects directly the whole performance of system. So far, variety real-time task allocation and scheduling algorithm each with advantages and disadvantages, have been made by domestic and foreign experts.This article presents a dynamic real-time parallel task allocating and scheduling algorithm based on Message-Oriented Middleware. It adapts a heterogeneous cluster-based structure and constructs a system model of dynamic real-time task parallel execution. This algorithm insures data security, stability and speed of data extraction by storing different partitioned data copies to different computation nodes. Task priority and node performance are two major factors considered by this algorithm when allocating and scheduling tasks through scheduling tool. Task priority is determined by task weight, worst running time, absolute deadline and the need of data resources. Node performance is determined by the type of computation node, distribution matrix and average computing velocity. By allocating each task to the most appropriate node, execution time will be minimum. And it can realize the dynamic allocating and scheduling to the tasks in the heterogeneous cluster mode and load balance. And rescheduling tasks on failed nodes insures success rate of task execution. As simulation results shows, this algorithm performs better than EDF algorithm and save algorithm in both execution time and execution success rate, which proves the algorithm is a stable and efficient task allocating and scheduling algorithm.