2012, © Copyright by authors, Published under agreement with IARIA- www.iaria.org The impact of workload on energy efficiency of virtualized systems

Abstract—Virtualization, i.e., running several virtual com-puters on the same physical hardware, is an essential tech-nology in data centers. Since demand for cloud computing services is constantly growing, an increasing number of data centers are focusing on improving their energy efficiency. This has made energy efficiency of virtualization technologies an important research domain. So far, maximizing performance of virtualization technologies has received a lot of attention in cloud computing industry and several academic studies on performance optimization can be found, too. However, these studies usually focus on improving energy efficiency by applying server consolidation methods. In this paper we focus on energy efficiency of virtualization technologies, i.e., how a virtual service can be made more energy efficient. Our aim is to reduce energy consumption without decreasing the quality of service. We have studied this by performing a large set of measurements with different system settings. We used both synthetic benchmarks and real applications. We found out that energy efficiency depends on 1) the workload of the virtual servers, and 2) the number of virtual servers on the physical server. We noticed that it is more energy efficient to maximize workload of virtual servers and to minimize their number. Additionally, we observed that properly configured idle virtual servers hardly increase energy consumption. Thus, our conclusion is that it is better to load virtual servers heavily or let them run idle. Keywords-virtualization; energy-efficiency; server consolida-tion; xen; kvm; invenio; cmss