Postcopy Live Migration with Guest-Cooperative Page Faults

Takahiro HIROFUCHI  Isaku YAMAHATA  Satoshi ITOH  

IEICE TRANSACTIONS on Information and Systems   Vol.E98-D   No.12   pp.2159-2167
Publication Date: 2015/12/01
Publicized: 2015/09/15
Online ISSN: 1745-1361
DOI: 10.1587/transinf.2015PAP0011
Type of Manuscript: Special Section PAPER (Special Section on Parallel and Distributed Computing and Networking)
Category: Operating System
virtual machine,  live migration,  page fault,  postcopy migration,  

Full Text: PDF>>
Buy this Article

Postcopy live migration is a promising alternative of virtual machine (VM) migration, which transfers memory pages after switching the execution host of a VM. It allows a shorter and more deterministic migration time than precopy migration. There is, however, a possibility that postcopy migration would degrade VM performance just after switching the execution host. In this paper, we propose a performance improvement technique of postcopy migration, extending the para-virtualized page fault mechanism of a virtual machine monitor. When the guest operating system accesses a not-yet-transferred memory page, our proposed mechanism allows the guest kernel to defer the execution of the current process until the page data is transferred. In parallel with the page transfer, the guest kernel can yield VCPU to other active processes. We implemented the proposed technique in our postcopy migration mechanism for Qemu/KVM. Through experiments, we confirmed that our technique successfully alleviated performance degradation of postcopy migration for web server and database benchmarks.