The Design And Implementation Of Image Location-aware And Caching System In Virtual Machine Cluster

With the development of virtualization technology,cloud computing,whose essence is relying on virtualization technology to build large-scale virtual machine cluster,is widely used in academic research and enterprise production for its flexibility,convenience,scalability and etc.Despite the rapid development of cloud computing,it is still in its early stages,and faces some problems such as too much deployment time of virtual machine,low efficiency of I/O in distributed storage and user experience issues.In this paper,we analysis and compare the key technologies related to the virtual machine cluster system,choose the construction scheme based on GlusterFS,QEMUKVM and OpenStack,and have done the following work:1)We designed two different architectures,separated or mixed for computing and storage,and then analysis and compare both architectures.2)We proposed the virtual machine scheduling optimization algorithm supporting image location-aware.The current OpenStack scheduling algorithm just relies on static physical resource information of all computing nodes,without considering the storage location of virtual machine image which has important influence.With the feature of image location-aware,we can eliminate the download process of virtual machine images,reduce much deployment time of virtual machine while providing high I/O efficiency.3)Image location-aware has its suitable scenarios,but for scenes that don't suit,we proposed the design of block caching and sharing system and implemented the corresponding LBCS virtual disk format to complete the function.It also reduces much deployment time of virtual machine while providing high I/O efficiency by the way of caching the data in distributed file system to local file system.The above features we proposed work coordinately to reduce the deployment time of virtual machine and also provide high I/O efficiency for running virtual machine instances,which makes the virtual machine cluster system more efficient and provide better experience for the usage of virtual machines.