Information Technology Reference
In-Depth Information
An Empirical Study of the Robustness of
Energy-Aware Schedulers for High Performance
Computing Systems under Uncertainty
Santiago Iturriaga, Sebastian Garcıa, and Sergio Nesmachnow
Universidad de la Republica
Montevideo, Uruguay
{siturria,sgarcia,sergion}@fing.edu.uy
Abstract.
This article presents an empirical evaluation of energy-aware
schedulers under uncertainties in both the execution time of tasks and
the energy consumption of the computing infrastructure. We address an
important problem with direct application in current clusters and dis-
tributed computing systems, by analyzing how the list scheduling tech-
niques proposed in a previous work behave when considering errors in
the execution time estimation of tasks and realistic deviations in the
power consumption. The experimental evaluation is performed over re-
alistic workloads and scenarios, and validated by in-situ measurements
using a power distribution unit. Results demonstrate that errors in real-
world scenarios have a significant impact on the accuracy of the schedul-
ing algorithms. Different online and oine scheduling approaches were
evaluated, and online approach showed improvements of up to 32% in
computing performance and up to 18% in energy consumption over the
oine approach using the same scheduling algorithm.
Keywords:
HPC, scheduling, energy-aware, uncertainty.
1 Introduction
Nowadays, energy eciency is a major concern when operating clusters, data-
centers, and grid/cloud computing infrastructures. From a global perspective,
all issues related to energy consumption raise several concerns for the scientific
community, including economic, environmental, and system performance [11].
Energy consumption on computing systems does not only depend on the
energy eciency and features of the hardware, but also on the software used
for task planning [1]. Among many different strategies for reducing the energy
consumption,energy-aware scheduling techniques have emerged as useful alter-
natives for accurate planning and lowering the power required for operation [16].
Energy reduction techniques are usually based on limiting the computing power
of the computing elements. They are in conflict with the system performance,
so applying them has an impact on the Quality of Service (QoS) perceived by
the user. Multi-objective formulations of the scheduling problem have been for-
mulated to account for the specific features of the trade-off between energy
utilization and performance [7].
