Research And Implementation Of High-performance Private Host Service Construction In Cloud Computing Environment

With the continuous development of cloud computing,cloud host relying on virtualization technology has been sought after by all walks of life.Users no longer purchase hardware service devices in advance as usual,but directly rent them on demand,which not only greatly satisfies the business construction,but also reduces the overall IT cost.The convenience is self-evident.The cloud host service provided by most cloud manufacturers is under the public cloud model,more to consider the basic needs of each user group,so the functions involved are complex,the overall structure is complex,covering a variety of technical ecology,which is not universal for some groups in specific environment.On the contrary,because of the access of cloud host service decentralizes the focus of work,the overall work efficiency is degraded.The purpose of this thesis is to build a set of high-performance private host service based on KVM virtualization technology in private environment and oriented to project development team.This thesis analyzes the requirements of private host service construction from the perspective of function and performance.Based on this,the highperformance private host service framework is designed in detail,and the four modules of host control,host allocation,host migration and image management are divided.The host control module completes the basic functional framework of private host service,realizes the authentication of user access and requests for services,sets up master-slave management nodes to ensure reliability,and cooperates with the host working nodes to complete the specific operation of private host.The heartbeat detection mechanism is used to monitor the host control nodes to ensure the healthy operation of the nodes.The host allocation module learns from the host scheduling strategy of OpenStack,optimizes and introduces the private host service in this thesis,realizes multi-factor dynamic weight scheduling strategy,comprehensively considering the load changes of multi-resource factors of the host node and the users' bias requirements for private host service,and provides users with a better host operating environment with overall performance.Host migration module adopts the joint judgment mechanism of double threshold trigger times to determine the best migration time for private hosts,selects the private hosts that match the current resource consumption of the host node for migration,and combines with the host allocation module,selects the appropriate host operating environment to move in,which greatly improves the stability of the overall service performance.The image management module mainly combines the distributed file system called MooseFS to manage the private host image template in a unified way,which provides convenience for users and avoids the endless spread of the images of each host node.In addition,the functions of each module of high-performance private host service are tested to confirm that the function of the module meets the anticipated requirements.At the same time,the performance test ensures that the private host system can meet the users' high-performance service requirements.