Virtualization technology play a vital role in cloud computing. In virtualization environment multiple applications can run at a same time. VM migration is one of the important features of the virtualization, which allows application to be transparently migrated along with their execution environments across physical machines. VM migration consists of four steps. These steps are source PM selection, VM selection from the selected PM in first step, target PM selection for placing the selected VM and selecting method for transfer VM data. In our proposed approach we are focusing on the last step i.e., choosing method for transferring VM data.
In this paper proposed approach is divided into two phases. First phase filter all pages which is modified in the last iteration and unmodified in the current iteration. In the second phase, page is divided into two types i.e., high dirty page and normal based on the modification in the last few iterations. For all filtered pages in the first phase now we check the number of times when the page is modified from the history record. If the page is modified more times the page will not be send in the current iteration otherwise the page will be send to the destination VM.
To evaluate the performance of the proposed approach it is implemented in CloudSim simulator and compare with the existing time series based pre-copy approach in term of total migration time and down time. Experiment result shows that proposed approach gives better result as compare to the base approach.