Abstract
Virtual machine live migration is a method of moving virtual machine across hosts within a virtualized datacenter. It provides significant benefits for administrator to manage datacenter efficiently. It reduces service interruption by transferring the virtual machine without stopping at source. Transfer of large number of virtual machine memory pages results in long migration time as well as downtime, which also affects the overall system performance. This situation becomes unbearable when migration takes place over slower network or a long distance migration within a cloud. In this paper, precopy based virtual machine live migration method is thoroughly analyzed to trace out the issues responsible for its performance drops. In order to address these issues, this paper proposes three phase optimization (TPO) method. It works in three phases as follows: (i) reduce the transfer of memory pages in first phase, (ii) reduce the transfer of duplicate pages by classifying frequently and non-frequently updated pages, and (iii) reduce the data sent in last iteration of migration by applying the simple RLE compression technique. As a result, each phase significantly reduces total pages transferred, total migration time and downtime respectively. The proposed TPO method is evaluated using different representative workloads on a Xen virtualized environment. Experimental results show that TPO method reduces total pages transferred by 71 %, total migration time by 70 %, downtime by 3 % for higher workload, and it does not impose significant overhead as compared to traditional precopy method. Comparison of TPO method with other methods is also done for supporting and showing its effectiveness. TPO method and precopy methods are also tested at different number of iterations. The TPO method gives better performance even with less number of iterations.
Highlights
Cloud Computing is considered as an utility based system for the dynamic provisioning of IT resources and services
This paper proposes a method, named as three phase optimization (TPO), which minimizes number of memory pages to be transferred in each iteration of the migration method
The analysis shows that the TPO further reduces total pages transferred, and total migration time for 10 iterations, for all types of workloads because the transfer of redundant pages in subsequent iterations is reduced
Summary
Cloud Computing is considered as an utility based system for the dynamic provisioning of IT resources and services. For providing on-demand and flexible provisioning of resources and services, it needs to utilize resources efficiently without any interruption due to maintenance and setup issues For this purpose, resources are shared among various users in such a way that the requirement of all users can be fulfilled. Resources are shared among various users in such a way that the requirement of all users can be fulfilled Virtualization makes it possible by running multiple operating systems and multiple applications on a single physical machine. It is mainly divided into three phases
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
