Evaluation of power consumption in programming models based on map reduce in shared memory systems

摘要

One of the most common models in parallel programming is Map reduce. In companies which use Map reduce framework, at each time a lot of computers are executing Map and Reduce functions. These functions are executing many times and we can estimate the effect of any small changes in response speed or consumed power in each execution of the Map reduce model to be very high. On the other hand, power and energy became an important challenge in computer systems with high performance, so that criteria such as consumed Power are as important as performance criteria. Nowadays, because of the generated heat and also because of decrease in energy sources, saving the consumed power is very important. So, finding the parts of the programs which needs more power is of prime Importance, because by finding these parts, we can find the ways to investigate and improves them in terms of power. In this paper, we investigate the available Map Reduce framework and programs in this regard from the point of consumed power, which are implemented in multi-core environment with common memory.