On Profiling the Energy Consumption of Distributed Simulations: A Case Study

Reducing the energy consumption of computing systems has been the topic of many research studies. Specifically, optimizing the energy consumption of applications is important to reduce the carbon footprint and the associated costs. In this paper, we detail a method to profile the energy usage of distributed simulations in a grid environment and conduct several experiments towards characterizing the power behavior of our PRIME network simulator. We conclude that although PRIME shows a high level of parallelization in some scenarios, the energy consumption increases significantly as more compute nodes are used to run a network model. Also, we show that cross-traffic has a tremendous impact in the energy usage.