基于杂交遗传算法的多处理器硬实时容错调度算法

传统的硬实时容错调度算法获得了较好的容错性能,但其任务拒绝率、处理器分配偏差比例以及最早完成时间等性能参数不佳,对此提出一种基于杂交遗传算法的优化方案,并对传统的硬实时容错算法进行优化。采用了中心型调度模型,并采用了任务备份方案来实现容错能力。将任务拒绝率、处理器分配偏差比例以及最早完成时间三个优化参数作为遗传算法适应度目标函数的三个带权分量,对其进行优化,通过遗传算法的杂交与迭代计算获得了优化的结果。最终使用不同的任务数量与处理器数量的组合对本算法与传统算法进行对比试验,结果可看出本算法的3个优化参数明显优于传统算法,且总适应度值亦比传统算法有明显改进。

Genetic algorithm based multiprocessor hard real time fault tolerant schedule

Traditional hard real-time fault tolerant scheduling algorithm has good fault tolerant performance, but its task rejection ration, processor allocation deviation ratio and earliest finishing time ratio performance is not good, aimed at that a optimal algorithm based on genetic algorithm is proposed, and the traditional hard real-time fault tolerant scheduling algorithm is optimized. The central scheduling model is used and the backup task approach is used to realize the fault tolerant. The three optimal parameters of task rejection ration, processor allocation deviation ratio and earliest finishing time ratio are set as the fit function of the genetic algorithm as well as the weight, the optimal results are got by crossover operation and iteration of genetic algorithm. Lastly, several set of different count of tasks and processors are set up for compare experimental. Three optimal parameters got from the proposed algorithm are better than the traditional scheduling algorithm, and the total fit value is better too.