Memory buddies: exploiting page sharing for smart colocation in virtualized data centers

  • Authors:

  • Timothy Wood;Gabriel Tarasuk-Levin;Prashant Shenoy;Peter Desnoyers;Emmanuel Cecchet;Mark D. Corner

  • Affiliations:

  • University of Massachusetts Amherst, Amherst, MA, USA;University of Massachusetts Amherst, Amherst, MA, USA;University of Massachusetts Amherst, Amherst, MA, USA;Northeastern University, Boston, USA;University of Massachusetts Amherst, Amherst, MA, USA;University of Massachusetts Amherst, Amherst, MA, USA

  • Venue:
  • Proceedings of the 2009 ACM SIGPLAN/SIGOPS international conference on Virtual execution environments
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Many data center virtualization solutions, such as VMware ESX, employ content-based page sharing to consolidate the resources of multiple servers. Page sharing identifies virtual machine memory pages with identical content and consolidates them into a single shared page. This technique, implemented at the host level, applies only between VMs placed on a given physical host. In a multi-server data center, opportunities for sharing may be lost because the VMs holding identical pages are resident on different hosts. In order to obtain the full benefit of content-based page sharing it is necessary to place virtual machines such that VMs with similar memory content are located on the same hosts. In this paper we present Memory Buddies, a memory sharing-aware placement system for virtual machines. This system includes a memory fingerprinting system to efficiently determine the sharing potential among a set of VMs, and compute more efficient placements. In addition it makes use of live migration to optimize VM placement as workloads change. We have implemented a prototype Memory Buddies system with VMware ESX Server and present experimental results on our testbed, as well as an analysis of an extensive memory trace study. Evaluation of our prototype using a mix of enterprise and e-commerce applications demonstrates an increase of data center capacity (i.e. number of VMs supported) of 17%, while imposing low overhead and scaling to as many as a thousand servers.