| The emergence of cloud computing brings great changes to the calculation. It can meet the needs of various fields to solve more complex problems, and deal with massive data. In order to efficiently use the lots of infrastructure resources in the cloud platform, virtualization technology is introduced into cloud computing, virtualization technology is one of the core technology of IaaS platform, because cloud computing platform cluster is distributed with a large variety of different services, and the load of each service in every moment is a lso different, with uncertainty. In order to make full use of these platform resources, it must carry load balancing on the entire platform resources. In cloud computing, live migration technology of virtual machine is an important method to realize load balancing.Traditional virtualization technology needs to run an additional complete operating system in a virtual machine, which is not only a waste of resources but also reduce the running efficiency of the virtual machine. The container technology is a good solution to the above problem, which does not need to run an additional complete operat ing system in a virtual machine. And each virtual machine shares the same kernel with its physical host. However, research on the live migration of the container technology is very small at present. This thesis is focusing on the study of the live migration mechanism of LXC containers by comparing with the traditional virtualization technology. The main contents are as follows:(1) This thesis first analyses the memory characteristics of LXC container, and proposes a live migration mechanism based on memory: AMC &T. The traditional pre-copy method of live migration mechanism requires synchronization on dirty pages frequently, which will lead to the transmission of large amounts of memory data, and repeatedly transfer a lot of redundant dirty page, so that the whole migration time will be too long. To solve the above problem, this thesis proposes a live migration mechanism for LXC container: AMC&T, which is based on adaptive memory compression algorithm: AMCA, and the AMC&T algorithm is validated by doing some experiments and analysis.(2) In view of migration mechanism of the non shared file system, this thesis analyzes the differences of virtual disk between LXC container and traditional virtual machine, and proposes a file system synchronization mechanism for container: FFSAS. FFSAS monitors and records changes of the LXC container to its file system, and these information will stored in the high speed cache in form of five tuples. After the LXC container is restored, its file system will be synchronized according to these information in high speed cache. And it is also an important guarantee to reduce data to be transferred during the whole migration process. After the LXC container is restored in the destination host, FFSAS will real time monitor the restored container ’s accesses to its file system, and finally realize the synchronization of dirty files of the entire file system of the LXC container. The migration mechanism proposed in this thesis for the non shared file system is actually a combination of FFSAS and AMC&T. In this thesis, above policy has been implemented and validated by doing some experiments and analysis. |
PDF Full Text Request