容器云多维资源利用率均衡调度研究  

Container cloud multi-dimensional resource utilization balanced scheduling

作　　者：龚坤 武永卫[1] 陈康[1] Gong Kun;Wu Yongwei;Chen Kang(Dept.of Computer Science&Technology,Tsinghua University,Beijing 100084,China)

机构地区：[1]清华大学计算机科学与技术系,北京100084

出　　处：《计算机应用研究》2020年第4期1102-1106,共5页Application Research of Computers

基　　金：国家重点研发计划资助项目(2016YFB1000504);国家自然科学基金资助项目(61433008,61373145,61572280,U1435216);中国博士后科学基金资助项目(2018M630162)。

摘　　要：在OpenShift容器云平台上针对其调度策略进行研究和改进,提出了基于多维资源空闲率权重的评价函数和调度方法。该方法综合考虑物理节点CPU、内存、磁盘、网络带宽空闲率和已部署的容器应用个数等因素,利用模糊层次分析法(fuzzy analytic hierarchy process,FAHP)自动建模求解容器应用多维资源权重参数。实验表明,新的调度方案能够使集群多维资源利用率更加均衡,从而提升资源的利用率和集群性能。This paper researched and improved the scheduling policy on the OpenShift container cloud platform and proposed the evaluation function and scheduling method based on the multi-dimensional resource idle rate weight.This scheme consi-dered the factors of CPU,memory,disk,bandwidth and the deployed container applications on the physical machines,and used the FAHP(fuzzy analytic hierarchy process)to automatically model and solve the multi-dimensional resource weighting parameters of the container applications.Experiments show that the new scheduling enables the cluster multi-dimensional resources utilization more balanced,improves resource utilization and cluster performance.

关 键 词：容器云 调度策略 OpenShift平台 评价函数 负载均衡 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]