Efficient pre copy virtual machine migration technique for cloud computing

Abstract

In a cloud computing platform, pre-copy Virtual Machine (VM) migration is the keytechnique that has revolutionized the working of data centers by addressing challengesof server consolidation, load balancing, system update, and hardware maintenance.In this technique, memory pages of the VM being migrated are iteratively copiedto the destination server while its execution at the source server continues. Efforts have been made by the researchers to improve the performance of the pre-copy VMmigration technique. However, the following issues still require special attention andneed to be addressed further. First, the pre-copy VM migration technique suffers froma page-level content redundancy problem, leading to performance penalties in terms ofan increase in downtime, migration time, and the amount of network traffic generated.Secondly, inaccurate fixed bandwidth allocation for a VM migration leads to performance degradation. Thirdly, there is a need to facilitate migration support for a new paradigm of virtualization technology called containerization. In the first contribution of this thesis, to address the page-level content redundancyproblem, a VM memory page content reduction technique has been proposed thattransfers the encoded form of dirtied pages. newlineThe second contribution of this thesis is to address the issue of inaccurate fixed bandwidth allocation by MIN and MAX schemes for a pre-copy VM migrationtechnique. The third contribution of this thesis is to address the issue of how to facilitatemigration support for container-based virtualization. Results obtained demonstrate that compared to the existing hypervisor-based pre-copy VM migration scheme, LXD/CRcontainer migration technique shows a significant improvement in terms of downtime, migration time, amount of data and the total number of pages transferred. newline newline