Power consumption and temperature measurement of virtualization solutions
Marius Marcu, Dacian Tudor, Sebastian Fuicu · International Workshop on Thermal Investigations of ICs and Systems · 2011
Modern data centers and cloud platforms demand higher levels of energy consumption and better temperature control together with the increasing processing demand and storage resources. Consolidation techniques using virtualization are currently the most promising methods used in today data centers to mitigate the effects of the increasing energy consumption trend. Although the false impression is that virtualization is the solution to most resource usage optimization problems encountered in data centers, the reality is that its usage raises additional challenges for system administrators and solution designers. One such specific challenge is to measure or at least to estimate the efficiency of a physical-virtual machines deployment. However, at the moment there is a lack of power and thermal metering for virtual machines (VMs). In this paper we evaluate several parameters that can be used to characterize energy consumption and thermal dissipation of a physical machine (PM) while hosting and executing several virtual machines. Furthermore, based on this evaluation we propose a test methodology that can be used to estimate energy efficiency and thermal efficiency of virtualized applications.