应用混合粒子群优化的检查点全局优化算法

The checkpoint global optimization algorithm based on the mixed particle swarm optimization

针对容错实时系统存在的局部最优检查点间隔为单次故障情况下的最优检查点间隔及局部最优检查点间隔并不是任务集全局最优检查点间隔的缺陷,首先给出检查点间隔全局优化问题的多目标优化模型,然后基于混合粒子群优化算法,提出检查点间隔全局优化算法.该算法通过混合粒子群优化算法的交叉和变异操作,避免算法陷入局部极值的困境,且增强了算法搜索全局近优检查点间隔的能力.实验表明,与其他检查点间隔优化算法相比,本算法可进一步提升系统容错能力.检查点间隔全局优化能在故障多次发生情况下,对任务集的检查点间隔进行全局搜索,以减小检查点设置次数和故障检测次数、高优先级任务抢占时间及故障恢复时间,提高系统可调度性.

For the task sets in the fault tolerant real time systems, the disadvantages of the local optimal checkpoint interval are under a single fault assumption and also not the global optimal checkpoint interval. To solve these, the multi-objective optimization model for the checkpoint interval global optimization was given first, and then the checkpoint interval global optimization algorithm based on the mixed particle swarm optimization algorithm was proposed. This algorithm avoids the shortcoming of falling into local optimum and enhances the ability of searching the global approximate optimal checkpoint interval by the crossover and mutation operations of the mixed particle swarm optimization algorithm, and further reduces the task worst case response time. The simulation results show that the algorithm can further improve the system fault resilience over the other checkpoint interval optimization algorithms. At the same time, the checkpoint interval global optimization can search the checkpoint intervals of the task set globally when the faults occur many times, by which the number of checkpoint and the number of fault detection can be reduced and the preemption time by the high priority tasks and the fault recovery time can also be decreased, and also the system schedulability can be improved.