A Data Persistence System For Virtual Ma-chines In Cloud Environment

Along with the vigorous development of virtualization technology, cloud computingplatforms emerge in endlessly. Cloud computing platforms meet user requirements of re-sources through the creation of virtual machines or virtual clusters. The data within virtualmachines is in the end store in the image file. This makes the user data tightly coupledwith the virtual machine image file, which is not conducive to data management, and thedata will be deleted once the virtual machine is shutdown.In view of the above questions, a data persistence system for virtual machines basedon cloud environment offers virtual machine independent block storage service. To avoidthe single point of failure problem and improve the system scalability, the system adoptsmaster-slave architecture and consists of one master node and several storage nodes. Themaster node is responsible for the user request processing; it uses heartbeat protocol tomaintain connections with other nodes. The storage node uses logical volume manager tomanage disk space and mount the disk to virtual machine through the small computer sys-tem interface. The allocation strategy chooses appropriate storage node to create block de-vice based on cluster network topology. Machines in the same rack are usually connectedto the same switch and racks connected together through another switch, resulting in treenetwork topology. For network communication within racks has low latency, the masternode chooses the storage node closest to the virtual machine to create block device.Function test results show that the virtual machine successful applies and mountsblock device. The data in the virtual machine is persistently stored in the block device.The performance test shows a higher access rate than the virtual image file, which meetsthe requirements of most applications.