摘要
容器云作为互联网底层基础服务设施应用越来越广泛,合理的任务调度对云资源优化、云服务质量以及企业的降本增效至关重要。为提高容器云环境下资源利用率,提出一种基于深度强化学习的资源调度算法(RLSD)。该算法基于深度强化学习理论构建动作空间、智能体状态等基本要素,通过采集各资源的利用率并结合权重因子实现奖励函数设计。通过仿真平台WorkflowSim构建交互环境,进行智能体的训练。实验表明,RLSD算法在保障任务调度稳定性的同时,资源利用率比传统算法提高了20%。
Container cloud is more and more widely used as the underlying infrastructure of the Internet.Reasonable task scheduling is crucial to cloud resource optimization,cloud service quality,and cost reduction and efficiency enhancement for enterprises.In order to improve resource utilization in container cloud environment,a resource scheduling algorithm based on deep reinforcement learning(RLSD)is proposed.The algorithm constructs basic elements such as action space and agent state based on deep reinforcement learning theory,and realizes reward function design by collecting the utilization rate of each resource and combining with weight factors.The interactive environment is constructed through the simulation platform WorkflowSim to train the agent.Experiments show that the RLSD algorithm can improve the resource utilization by 20%compared with the traditional algorithm while ensuring the stability of task scheduling.
作者
俞延峰
孙雯雯
陈雷放
YU Yanfeng;SUN Wenwen;CHEN Leifang(College of Computer Science and Technology,China University of Petroleum,Qingdao 266520,China;Information Engineering Department,Shandong Foreign Trade Vocational College,Qingdao 266100,China;School of science and information,Qingdao Agricultural University,Qingdao 266109,China)
出处
《电子设计工程》
2023年第10期59-63,68,共6页
Electronic Design Engineering
基金
山东省自然科学基金面上项目(ZR2019MF049)。
关键词
云计算
容器
强化学习
资源调度
cloud computing
container
reinforcement learning
resource scheduling