| Virtualization technology can break the original physical state of the hardware resources, and redistribution to meet the needs of different users on computing resources in a shared cloud platform environment. In reality, there is a great difference of requirement to the computing resource between different characteristics of the application, and a large number of memory pages changed frequently will be transferred as the migration of the virtual machine, resulting in the bad performance of the applications and the live migration of virtual machine, which have a bad effect on performance of the cloud platform. Therefore, the virtual machine live migration is one of the hottest research in cloud computing.Currently, the widely recognized Pre-copy migration based on probability prediction does have better performance in the case of the stable dirty pages rate. However, due to the different businesses require different resources at different time, there are various ways in the memory access. When the dirty page rate changes suddenly, the probability forecast will fail intermittently, and a large number of data will be repeated transmission, which affect the migration performance. Common compression algorithm can compress the memory data, reducing the amount of data transferred, shorter the transmission time. The different compression algorithms have different performance to deal with the different characteristics of the data, which is taken into account, but did not take into account applications can cause the memory change constantly. And the compression algorithm does not make an appropriate adjustment for it.Therefore, on the basis of the migration framework of Pre-copy migration based on probability, this paper makes an improvement on it. Two sub modules are added into the transfer execution module, and one is an adaptive bandwidth module based on business characteristics, the other is a mixed compression module based on the data characteristics. The adaptive bandwidth module can adjust the transmission bandwidth according to the characteristics of business, which shorten each iteration time of transmission, and make a virtuous circle. At the same time, according to the data characteristic, the different compression is used to compress data, which shorten the total migration time and downtime, and improve the migration performance.In order to test the rationality and feasibility of the migration strategy proposed by this article, some special experiment test platform and environment are set up. The experimental data shows that the migration strategy proposed by this article can reduce the total migration time and downtime in the environment with steady and unsteady high dirty page, which can improve the migration performance. |
PDF Full Text Request