云环境下基于强化学习的多目标任务调度算法

针对云计算环境下的多目标任务调度问题,提出一种新的基于Q学习的多目标优化任务调度算法(Multi-objective Task Scheduling Algorithm based on Q-learning,QM TS).该算法的主要思想是:首先,在任务排序阶段利用Q-learning算法中的自学习过程得到更加合理的任务序列;然后,在虚拟机分配阶段使用线性加权法综合考虑任务最早完成时间和计算节点的计算成本,达到同时优化多目标问题的目的;最后,以产生更小的makespan和总成本为目标函数对任务进行调度,得到任务完成后的实验结果.实验结果表明,QMTS算法在使用Q-learning对任务进行排序后可以得到比HEFT算法更小的makespan;并且根据优化多目标调度策略在任务执行过程中减少了makespan和总成本,是一种有效的多目标优化任务调度算法.

Multi-objective Task Scheduling Algorithm Based on Reinforcement Learning in Cloud Environments

In order to solve the multi-objective task scheduling problem in the cloud environment,a novel multi-objective optimization task scheduling algorithm based on Q-learning( Multi-objective Task Scheduling Algorithm based on Q-learning,QMTS) is proposed.The main idea of the QMTS algorithm is: firstly,in the task ordering phase,uses the self-learning process in the Q-learning algorithm to obtain a more reasonable task order;secondly,in the virtual machine allocation phase,uses the linear weighting method to consider the task’s earliest finish time and computation cost of computing nodes to optimize multi-objective task scheduling problems;finally,the task is scheduled with the objective function of generating smaller makespan and total cost,and the experimental results are obtained after tasks are completed. The experimental results show that QMTS algorithm can obtain a smaller makespan after sorting tasks by using Q-learning;and it reduces the makespan and total cost in the task execution process according to the optimized multi-objective scheduling strategy. QMTS algorithm is an effective multi-objective optimization task scheduling algorithm.