Survey on Techniques used to Measure Power and Energy Consumption

In today's world of big data where we moving up from terascale computing to petascale and approaching exascale computing. Power consumption in such large scale computing is a critical concern. We need to measure power and energy consumption to optimize the power consumed without affecting the performance of the application. Most of the scientific applications use HPC architecture for computation, these applications consume large amount of power. As the power consumed is more we need some sort of tools and techniques to reduce the amount of power consumption. We understand the importance of energy efficiency and power optimization. To optimize the power consumption we first need to measure power and energy usage. The survey mainly focuses on different techniques used in calculating and measuring power and energy consumed by software applications. Applications run on different architectures need to be dynamically monitored. The methodologies can be adopted on different processors to predict power usage. The portability of these techniques is achieved through PMU’s (Performance monitoring units). Many of the existing solutions make use of external wattmeter or hardware meters. To optimize power and cost, it is better to use software tools rather than any other hardware meters, as additional cost of hardware can be reduced. This in turn will reduce the operational cost, as the additional meters will use a certain amount of power for its own functionality. The study also highlights the fact that most of the research work is done on Linux platform, as Linux platforms are open source it is easier for researchers to avail all the possible operating system support through packages and other related details, however this is scalable to other platforms.