Laurent Lefèvre

Your cluster is not power homogeneous: Take care when designing green schedulers!

Future supercomputers will consume enormous amounts of energy. These very large scale systems will gather many homogeneous clusters. In this paper, we analyze the power consumption of the nodes from different homogeneous clusters during different workloads. We classically observe that these nodes exhibit the same level of performance. But we also show that different nodes from a homogeneous cluster may exhibit heterogeneous idle power energy consumption even if they are made of identical hardware. Hence, we propose an experimental methodology to understand such differences. We show that CPUs are responsible for such heterogeneity which can reach 20% in terms of energy consumption. So energy aware (Green) schedulers must take care of such hidden heterogeneity in order to propose efficient mapping of tasks. To consume less energy, we propose an energy-aware scheduling approach taking into account the heterogeneous idle power consumption of homogeneous nodes. It shows that we are able to save energy up to 17% while exploiting the high power heterogeneity that may exist in some homogeneous clusters.

Energy-Efficient Data Transfers in Large-Scale Distributed Systems

Data centers are becoming one of the major consumers of energy in modern economies accounting for a measurable fraction of the total energy consumption. Furthermore, the energy consumption of data centers has a high growth rate with the energy consumption doubling about every six years. With large data centers consuming several MW of power or even 10’s of MW energy efficiency and energy consumption of data centers have become of great importance economically and environmentally. Up to a decade or so ago it was common that the computing equipment in data centers only consumed a third or less of the total energy consumption. Today, in well-designed data centers the computing equipment consumes at least 60 – 70% of the total power, and in some designs as much as 90% or more of the total power. To further enhance energy efficiency various forms of energy reuse is also pursued thereby reducing the energy consumption for other purposes and hence also contributing to reduced emissions. Reducing the energy consumption for computing has for the last decade also been one of the key drivers for CPU and platform vendors. Increasingly, energy efficient computing system and data center design requires an integrated approach to server and system design, cooling and potential energy reuse. In this chapter we will review recent developments as well as emerging technologies for energy efficient server and data center designs, as well as the challenges in using clean energy sources.