An enhanced approach for estimating writable working set by pre-copy and prediction

Because cloud computing and cloud-based applications have attracted more and more attention in recent years, more and more servers are needed to provide services. However, each traditional server can execute only one application at one time. In order to satisfy user requirements and increase the server utilization rates, server virtualization has thus presented to the public. For clients to experience seamless services, live migration has become a significant research topic. Previous researches have revealed that two important parameters affecting live migration are downtime and total migration time. Traditional pre-copy approaches can reduce the downtime but the total migration time is too long due to repeated retransmissions. Therefore, to decrease the downtime and total migration time, our proposed scheme sets the threshold values based on memory modification prediction technique to determine whether the pages must be transferred in the stop-and-copy phase.