Estimation by Software for the Power Consumption of Streaming-Media Servers

The power consumption of a streaming-media server can be obtained in real time by using the virtual-instrumentation software described in this paper without using an additional hardware meter. We have built a model to estimate the power consumption from the observation of experimental data that consists of tracking the CPU utilization and detecting the operation parameters of the measured servers. When calculated in real time, the CPU utilization can respond to the dynamic change of the power consumption of the measured servers. The operation parameters represent the hardware configuration of the measured servers. We also propose three methods to obtain these parameters: filled- manually, hardware-revised, and software-revised parameter estimating. We have constructed the virtual-instrumentation software according to this power model to measure the power consumption of the streaming-media server. To facilitate the measurement process, we have also designed a suitable graphic-user interface for it. Our virtual-instrumentation software with three parameter-estimating methods has been tested by way of comparison with measurement results obtained by a power meter. The average power values of the hardware-revised method are found to yield mean errors of the estimate within 3%. The mean error of the software-revised method is within 6%. However, the filled- manually method may underestimate the power consumption by as much as 11%.