System Power Model and Virtual Machine Power Metering for Cloud Computing Pricing
Chengjian Wen, Xiang Long, Yang Yang, Fan Ni, Yifen Mu · 2013
Multi-core virtualization platforms has been the basic infrastructure for data center for which Green computing and cloud computing are the most significant trends. Most servers don't have build-in power measurement sensors in modern data center. Besides, even if the total server power can be measured in real time VM(virtual machine) power cannot be measured purely by any power sensor. A suitable VM power model can help data center operator save power and price the VM energy consumption in cloud computing platforms. We present a solution for system power estimation and VM power metering by using performance events counter. We build power models to infer power consumption from the system resource usage such as cpu and memory which can be indicated by certain performance events counter value. The result shows that this method can get the accuracy of 97% on average.