Research And Implementation Of Highly Available Cloud Service

With the rapid emergence and rise of cloud computing,the IT industry has developed rapidly,and an increasing number of IT services in traditional enterprises and institutions are developing in a more intelligent,convenient and efficient direction.The idea of cloud computing affects more and more researchers as well.The core of which is to put all computing services in the cloud.By taking full advantage of the idle physical hardware resources,cloud computing could meet more business requirements,completing unified management of resources and centralized data placement.It compensates for the difficulty of the resources waste and management in the traditional IT industry.One core technology of cloud computing is to implement the virtualization of all hardware resources.Consequently,virtualization technology has became a subject that major technology vendors are scrambling to study.From the implementation of the underlying virtualization,such as KVM,Xen,HyperV and other technologies to virtualization management platforms such as OpenStack and CloudStack,the field of virtualization is gradually enriched.To summarize,the virtualization technology of hardware resource performs a particularly important role in cloud computing.The research background of this thesis is the virtualization platform developed by the laboratory.By analyzing the existing function the author found out that there are still many problems to be solved,such as unavoidable single machine failure,lack of image management,load monitoring mechanism and backup solution.Based on the analysis above,the author deeply studied the KVM virtualization technology,the open source Ceph storage system and the currently popular OpenStack platform.After learning these core concepts and design ideas,the author proposed a brand-new architecture for the current virtualization platform.This architecture is designed to achieve high availability of platform services.In the design of this architecture,firstly,the host and the virtual machine disk image are separated,and each host is regarded as a computing node.These nodes compose a computing cluster to avoid single-machine operation.Then base on Ceph a distributed storage cluster is built to realize the ability of shared storage and backup for the mirror and disk of the virtual machine.At the same time,the author researched the schema,running the virtual machine on the Ceph RBD block device,to simplify the work of the computing node.Finally,the thesis designed a cluster monitoring module and the deviation correction prediction algorithm based on the integrated load to schedule the virtual machines,which realizes the automatic load balancing of this system.The test results demonstrated that the high-availability virtualization platform designed and implemented in this thesis provides stable and reliable virtual machine services.The flexible virtual machine migration mechanism and image storage solution ensure the efficiency and stability of this platform.The virtualization platform with the new architecture is divided into four modules: storage cluster,computing cluster,management center and monitoring center,and it reduces the coupling of the system,which is favorable for the extension of platform and subsequent research.