基于稳定匹配的容器部署策略的优化

Docker的发展使得操作系统级虚拟化的容器渐渐兴起,容器即服务(CaaS)也越来越普及。随着容器技术的发展,容器将成为云环境中的主要部署模型,但针对容器的整合部署技术还未得到广泛的研究。容器化云环境中的容器数量众多,如何将众多的容器部署到合适的虚拟机以降低 数据中心能耗,成为了一个亟待解决的问题。因此,文中创新性地将机器学习中的几种相似度计算方法作为稳定匹配算法的偏好规则,同时将已经拟分配过容器的虚拟机继续加入偏好列表,从而将一对一的稳定婚姻匹配算法改进为多对一的稳定匹配,解决了将容器整合到虚拟机上的初始化部署问题。仿真实验结果表明 , 采用优化的稳定匹配算法来初始化将部署容器时,不仅SLA违规较低,而且比FirstFit,MostFull以及Random算法分别约节能12.8%,34.6%和30.87%,其中使用欧氏距离作为稳定匹配算法偏好规则的节能效果最好。

Optimization of Container Deployment Strategy Based on Stable Matching

With the development of Docker,virtualization containers of operating system level are on the rise,and contai-ner-as-a-service(CaaS) is also becoming more and more popular.With the development of container technology,the container will become the main deployment model in the cloud environment,but the integrated deployment technology for the container has not been widely studied.In the cloud environment,how to deploy a large number of containers to a suitable virtual machine to reduce the energy consumption of the data center becomes an problem which needs to to be solved urgently.Therefore,several similarity calculation methods in machine learning are used as the preference rules of the stabilization matching algorithm,and the virtual machines that have been allocated to the container are added to the preference list at the same time,which makes the one-to-one stable marriage matching algorithm update to many-to-one stable match,solving the initial deployment problem of integrate container into the virtual machine.The experimental results show that the optimal storage efficiency is about 12.8%,34.6% and 30.87% compared with the FirstFit,MostFull and Random methods respectively,and when the Euclidean distance is used as the preference rules of stabilization matching algorithm,the performance of energy saving is the best.