Modeling energy and performance of Phoenix++ based parallel programs

In order to provide more processing power, computer systems manufacturers are trying to increase the performance of their products. Nowadays, multi-core processors are being used as a solution to reach more performance via thread level parallelism. MapReduce are considered as one of the most appropriate models of programming for data parallelism. Different frameworks have been developed based on this model. Phoenix++ is one of these frameworks that has been implemented for shared memory systems. Since computing is becoming more and more important in people's life, more and more energy spent computation. Energy is also becoming as big challenge in human life. Therefore doing energy efficient computation is becoming very important. For a long time, performance was considered as the most important metric for computing systems. However energy consumption has become as important as performance these days. We offer models in this paper to determine a system configuration for parallel programs based on Phoenix++ framework in way to minimize energy delay product (EDP) by selecting proper number of active cores, the number of threads, and working frequency level. The model accuracy is evaluated by running experiments on a real system.