Home // CLOUD COMPUTING 2012, The Third International Conference on Cloud Computing, GRIDs, and Virtualization // View article

Maximizing Utilization in Private IaaS Clouds with Heterogenous Load

Authors:
Tomas Vondra
Jan Sedivy

Keywords: Cloud Computing; Automatic Scaling; Job Scheduling; Real-time Infrastucture

Abstract:
This document presents ongoing work on creating a computing system that can run two types of workloads on a private cloud computing cluster, namely web servers and batch computing jobs, in a way that would maximize utilization of the computing infrastructure. The idea stems from the experience with the Eucalyptus private cloud system, which is used for cloud research at the Dept. of Cybernetics. This cloud lets researchers use spare computing power of lab computers with the help of our in-house queue engine called Cloud Gunther. This application improves upon current practices of running batch computations in the cloud by integrating control of virtual machine provisioning within the job scheduler. In contrast to other similar systems, it was built with the capacity restrictions of private clouds in mind. The Eucalyptus system has also been evaluated for web server use, and the possibility of dynamically changing the number of servers depending on user demand, which changes throughout the day, has been validated. Although there are already tools for running interactive services in the cloud and tools for batch workloads, there is no tool that would be able to efficiently distribute resources between these two in private cloud computing environments. Therefore, it is difficult for the owners of private clouds to fully exploit the potential of running heterogenous load while keeping the utilization of the servers at optimal levels. The Cloud Gunther application will be modified to monitor the resource consumption of interactive traffic in time and use that information to efficiently fill the remaining capacity with its batch jobs, therefore raising the utilization of the cluster without disrupting interactive traffic.

Pages: 169 to 173

Copyright: Copyright (c) IARIA, 2012

Publication date: July 22, 2012

Published in: conference

ISSN: 2308-4294

ISBN: 978-1-61208-216-5

Location: Nice, France

Dates: from July 22, 2012 to July 27, 2012