Software Based DFS Technique for Parallel Applications to Conserve the Power

Scientific parallel applications need more computing power, to produce such computing power we need high end hardware. This hardware consumes more power and generates lot of heat. Cooling system is required to make sure that the hardware is healthy otherwise it will be prone to failure. As a result, cost of data center operation will be more and it will not be environmental friendly. By reducing the power consumption we can reduce the cost as well as make data center environmental friendly. This work proposes software based Dynamic Frequency Scaling technique which helps in reducing the power consumption by parallel applications. To evaluate our technique we used popular NAS parallel benchmark and observed that up to 40% energy saving is achieved.