A Study on Performance of Processes in Migrating Virtual Machines

Yosuke Kuno, Kenichi Nii, Saneyasu Yamaguchi

2011 · 35 citations · 7 references

Concepts

TL;DR

Virtual machine migration in cloud environments occurs via non‑live or live methods, the former halting processes and the latter allowing continuous operation but potentially degrading performance. The study aims to determine which migration method yields better process performance. The authors evaluate both methods by measuring process performance during migration, showing severe decline in live migration. Analysis reveals that host‑OS communication and memory writes are key contributors to performance degradation.

Abstract

In a cloud computing environment, virtual machines are migrated with two kind of methods. One is non-live migration, and the other is live migration. In case of non-live migration, a virtual machine stops their processes during migrations. In case of live migration, a virtual machine and its processes keep running during migration, but their performance may severely decrease. In order to obtain better performance, the suitable migration method should be chosen. In this paper, we present performance evaluation of both migration methods, and demonstrate that performance of processes on a migrating virtual machine severely declines. The analysis for this degradation is presented, and it is revealed that a host OS communication and memory writing are important reasons for the decline.

References

7