In this paper, we analytically model the behavior of gigabit Energy Efficient Ethernet (EEE) links with coalescing using M/G/1 queues with sleep and wake-up periods. The particularity of gigabit EEE links is that energy-saving operations are triggered only when links are inactive in both transmission directions. Our model approximates with a good accuracy both the energy saving and the average packet delay by using a few significant traffic descriptors. Furthermore, we use real traffic traces to investigate on the performance of static as well as dynamic coalescing schemes. Surprisingly, our evaluation shows that dynamic coalescing does not significantly outperform static coalescing.

In this work, we propose a dynamic coalescing algorithm for IEEE 802.3az standard, which dynamically adapts coalescing operations based on the current load and on the delay experienced in the link. Our results show that our algorithm almost doubles the energy efficiency of EEE with static coalescing while keeping packet delay bounded.

In this work, we present an exhaustive empirical characterization of the power requirements of multiple components of data center servers. To do so, we devise different experiments to stress these components, taking into account the multiple available frequencies and the fact that we are working with multicore servers. In these experiments, we measure energy consumption of server components and identify their optimal operational points. Our study proves that the curve defining the minimal CPU power utilization, as a function of the load in active cycles per second, is neither concave nor purely convex. Instead, it definitively shows a super-linear dependence on the load. Similarly, we present results on how to improve the efficiency of network cards and disks. Finally, we validate the accuracy of the model derived from our characterization by comparing the real energy consumed by two Hadoop applications-PageRank and WordCount-with the estimation from our model, obtaining errors below 4.1% on average.