Whole-system live migration mechanism for virtual machines

We describe a whole-system live migration mechanism which transfers the whole run-time state including local disk storage, in this paper. A three-phase migration (TPM) algorithm is proposed to minimize the downtime caused by migrating large disk storage data and keep data integrity and consistency. A Block-bitmap is used to track all the write accesses to the local disk storage during the migration, and then to direct synchronization. Also, an incremental migration (IM) algorithm is described to facilitate the migration back to initial source machine. Our experimental study based on Xen shows that the algorithms perform well even when I/O-intensive workloads run in the migrated VM. The downtime of whole-system migration is close to shared-storage migration algorithms. The Block-Bitmap based synchronization mechanism used in both TPM and IM is simple and effective, and the performance overhead of recording all the writes is less than 1 percent.EI04894-8993