Home // International Journal On Advances in Systems and Measurements, volume 5, numbers 1 and 2, 2012 // View article

Dependable Estimation of Downtime for Virtual Machine Live Migration

Authors:
Felix Salfner
Peter Tröger
Matthias Richly

Keywords: virtual machine; live migration; downtime analysis;

Abstract:
Modern virtualization environments allow the live migration of running systems for load balancing and failover purposes in case of failing hosts. The overall duration of such migration and the short downtime during this process are essential properties when implementing service availability agreements. However, both metrics are currently only deter- minable through direct experimentation. For this reason, we present a new model for estimating the worst-case values of migration time and downtime in live migration. The prediction is based on a small set of input parameters characterizing the application load and the behavior of the host. We performed a large set of experimental evaluations for the model with three different virtualization products. The results show that total migration time as well as downtime are mainly influenced by the memory utilization pattern inside the virtualized system. The experiments also confirm that the proposed model can pre- dict worst-case live migration performance with high accuracy, rendering the model a useful tool for implementing proactive virtual machine migration.

Pages: 70 to 88

Copyright: Copyright (c) to authors, 2012. Used with permission.

Publication date: June 30, 2012

Published in: journal

ISSN: 1942-261x