Research On Affinity,Security And Memory Efficiency Of Cloud Platform Scheduling

Cloud computing is an important cornerstone of modern Internet services.Cloud computing provides services in the form of a cloud platform,and scheduling is a key component of the cloud platform system.The cloud platform abstracts the resources such as hardware,software and service into resource pool by virtualization technology,and manages the resources by the scheduling system.Application service performance,resource utilization rate and security protection are important objectives for the design of cloud platform scheduling system.According to the principle of virtualization,there is a certain correlation between the virtual machines,named affinity.The different deployment of virtual machines will affect the application performance of the virtual machine and the utilization rate of cloud platform resources.In OpenStack cloud platform,the scheduling algorithm is too simple,and it lacks a mechanism to make use of these phenomena.On the other hand,in cloud platform scheduling system,important configuration parameters are stored in memory key-value storage service,while traditional memory protection methods can not prevent hijack operation system from illegal access,and there is a certain security risk.When the scheduling results lead to a centralized deployment of virtual machines,there will be a lot of free memory on some servers,and the memory efficiency of the cloud platform is low.In order to design a reasonable,safe and efficient cloud platform scheduling system,this paper focuses on the cloud platform scheduling problem,and studies the 3 aspects of cloud platform scheduling system,including affinity,security protection and memory efficiency.The main contributions are:1.Study on the virtual machine affinity.This paper analyzes the influence of different deployment methods on the performance of virtual machine application and the utilization rate of cloud platform resources,and defines affinity to describe the relationship between virtual machines.We design and implement an OpenStack cloud platform scheduling based on virtual machine affinity to improve application performance and resource utilization ratio.2.Study on the memory security protection.This paper designs and implements the memory key-value storage service based on SGX hardware protection,which is used for the important configuration parameters of the security storage scheduling system.We analyze the consumption hotspots between modules,and improve the read and write performance by batch optimization.3.Study on the memory usage efficiency.In order to cope with the excess memory resources caused by scheduling results,this paper designs and implements the Docker remote memory volume management based on SGX key-value service.We use remote memory to improve the IO performance of container applications and improve the memory efficiency of cloud platform.