Research And Improvement Of Virtual Machine Dynamic Migration Based On Xen In Cloud Computing

Calculation model has been changing, and the fundamental reason is therapid development of computer hardware and software technology, and the cloudcomputing model generated in this technical background. It uses an innovativebusiness model, so users use computing services freely and paies according to theamount. With virtualization scalability and flexibility, cloud computing improvesresource utilization, simplifies management and maintenance of resources andservices. With information services automation technology, cloud computing canprovide packaged resources available to users. In the cloud computing model,resources demand of users is always changing. To avoid hot spots (hotsPot),when users obtian resources, cloud platform needs balance load between serversdynamically. Reserve resource too early is a waste of resources, cloud computingneeds to solve these problems urgently.To solve the above problems, virtualization technology offers a solution.Service is packaged in a virtual machine, the packaged service is mapped to aphysical server. Dynamic migration technology of virtual machine can achieveremapping, running virtual machines can be migrated from one machine toanother completely and seamlessly. So that users almost can not feel migrateprocess, and it will not affect the services provided externally.In order to reduce downtime and total migration time, Xen virtual machinesuses traditional Pre-Copy algorithm to migrate, Pre-Copy algorithm perform wellin empty load and low load, but not perform very well in high load. In this paper,improvement of Xen virtual machine live migration technology is studied. Thedynamic migration framework and migration algorithm are improved, predictionprobability of modules and compressed memory modules have been added to original dynamic migration mechanism base on in-depth study of original livemigration framework and pre-copy algorithm of Xen. Markov Model is used tothe prediction probability of modules, prediction probability algorithm isproposed. Memory compression algorithm is presented in memory compressionmodule, the algorithm compresses memory page of migration domain, reducedthe amount of migration data, reduced burden network. Data structure designideas of predicted probability algorithm and memory compression algorithm aredescribed. The experiment results of this paper show that the improvedmechanism and algorithm can effectively improve migration performance.