Efficient and Flexible Checkpoint/Restore of Split-memory Virtual Machines

Abstract

Recently, clouds provide virtual machines (VMs) with a large amount of memory for big data analysis. For easier migration of such VMs, split migration divides the memory of a VM into several pieces and transfers them to multiple hosts. Since the migrated VM called a split-memory VM needs to exchange memory data between the hosts, it is inherently subject to host and network failures. As a countermeasure, a checkpoint/restore mechanism has been used to periodically save the state of a VM, but the traditional mechanism is not suitable for split-memory VMs. It has to move a large amount of memory data between hosts during checkpointing and can just restores a normal VM on one host. This paper proposes D-CRES for efficient and flexible checkpoint/restore of split-memory VMs. D-CRES achieves fast checkpointing by saving the memory of a VM in parallel at all the hosts without moving memory data. For live checkpointing, it consistently saves the memory of a running VM by considering memory data exchanged by the VM itself. In addition, it enables a split-memory VM to be restored in parallel at multiple hosts. We have implemented checkpoint/restore of D-CRES in KVM and showed that the performance was up to 5.4 times higher than that of using the traditional mechanism.