Jandaghi, S.J. and Bhattacharyya, A. and Sotiriadis, Stelios and Amza, C. (2016) Consolidation of underutilized virtual machines to reduce total power usage. In: UNSPECIFIED (ed.) CASCON '16 Proceedings of the 26th Annual International Conference on Computer Science and Software Engineering. Riverton, U.S.: IBM Corp., pp. 128-137.
![[img]](https://eprints.bbk.ac.uk/style/images/fileicons/text.png) |
Text
p128-jandaghi.pdf_ip=193.61.44.35&id=3049890&acc=ACTIVE SERVICE&key=BF07A2EE685417C5.5906F0ED3F2D94AB.4D4702B0C3E38B35.4D4702B0C3E38B35&__acm__=1521649525_39ddb7b2ec2177cfc21d7d7992e88b1c - Published Version of Record Restricted to Repository staff only Download (1MB) | Request a copy |
Abstract
Data centers (DC) have become one of the biggest markets with new challenges and opportunities in the past decade. Many big companies are owners of DCs providing services and cloud solutions. With this increasing demand and in- terest in cloud service, thousands of virtual machines (VM) are being instantiated to run a variety of services in a data center. Beside the benefit of service provisioning, a DC is a big consumer of electric power and producer of greenhouse gasses consequently. Because the VMs are not using their assigned resources all the time, resources are multiplexed via virtualization. However, current resource management methods are oblivious to actual utilization or power con- sumption. Monitoring of servers in big data centers like Google and Twitter has shown that the current resource utilization is less than fifty percent in total. Specifically, in this paper, we use OpenStack, a popu- lar cloud management software to orchestrate and manage VMs. In Open-Stack, the virtualization factor is a constant value oblivious to VM resource consumption. We design and implement two dynamic methods for adapting the vir- tualization factor based on the monitored VM resource con- sumption. In our first method, we identify VMs that are mostly idle, and we opportunistically migrate all idle VMs to one or more servers in such a way to keep the chance of performance degradation to a minimum, while saving total resources. In our second, more general method, we model consolidation of underutilized VMs as a knapsack problem. We show that our methods save a significant amount of un- derutilized resources while minimizing performance degra- dation during and after dynamic reconfiguration.
Metadata
| Item Type: | Book Section |
|---|---|
| Keyword(s) / Subject(s): | cloud computing, data centers, utilization |
| School: | Birkbeck Faculties and Schools > Faculty of Science > School of Computing and Mathematical Sciences |
| Research Centres and Institutes: | Birkbeck Knowledge Lab |
| Depositing User: | Stelios Sotiriadis |
| Date Deposited: | 02 May 2018 14:26 |
| Last Modified: | 09 Aug 2023 12:43 |
| URI: | https://eprints.bbk.ac.uk/id/eprint/21805 |
Statistics
Additional statistics are available via IRStats2.
Archive Staff Only (login required)
