摘要
为了提高云环境的可靠性,对虚拟资源的管理是一个关键.针对虚拟机的自适应配置问题,提出一种分层强化学习Options和蚁群优化算法融合的方法 A-HRL.该方法记录蚂蚁遍历过程中留下的信息素,利用信息素变化率引入粗糙度概念,并根据粗糙度阈值创建子目标实现任务分层.将A-HRL算法应用于虚拟机自适应配置中,通过创建任务组和虚拟机可用性评估表监督每个任务的进度与质量,最大限度地提高每个应用的性能.实验结果表明,A-HRL算法比传统的强化学习算法性能更优.
In order to improve the reliability of the cloud, the management of the virtual resource is a key. According to virtual machine adaptive allocation problem, this paper proposes a methods of A-HRL, which fuses the hierarchical reinforcement learning Options and the ant colony optimization algorithm. Ants leave pheromone during the traversal,and pheromones change rate is introduced into the concept of roughness. And creates the hierarchical sub-goals according to the threshold of roughness, thus able to explore more ef- fective. A-HRL algorithm is applied to adaptive applicatioas in the virtual machine configuration,this method creates a task group and virtual machine availability assessment, supervising each task progress and quality, and maximizes performance for each application. The algorithm is used to analyse the virtual machine configuration in CloudSim simulation, and the results show that this algorithm has better performance than the RL algorithm.
出处
《小型微型计算机系统》
CSCD
北大核心
2015年第4期801-805,共5页
Journal of Chinese Computer Systems
基金
云计算中虚拟机资源与应用系统参数的协同自适应配置研究基金项目(61272382)资助
关键词
OPTIONS
蚁群算法
虚拟机
自适应配置
Options
ant colony optimization algorithm
virtual machine
adaptive configuration
