Working Set Prediction using LRU cache with Splay Tree Algorithm in Live VM Migration

Live virtual machine (VM) migrationprovides great benefits for load balancing,power management, fault tolerance and othersystem maintenance issues in modern datacenters. VM live migration basically consists oftransferring its memory pages from a sourceserver to a destination server. The amount oftransferred memory pages affected the downtimeand application performance of virtualmachines. In pre-copy approach in live VMmigration, total migration time is prolonged thesignificant amount of transferred data during thewhole migration process. In this paper, wepropose a working set prediction algorithmwhich combine LRU (Least Recently Used) cachewith splay tree algorithm. Deploying splay treealgorithm with LRU, the proposed algorithmpredict the memory pages of the most recentprocess and define as the working set. Applyingthe proposed algorithm, we can reduce theamount of transferred memory pages duringmigration process by transferring the workingset in last round. Experiment demonstrates thatcompared with XEN’s default pre-copy basedmigration algorithm, the proposed algorithm canreduce 23.67% of the total transferred memorypages during migration process