Information Technology Reference
In-Depth Information
Multiobjective Energy-Aware Datacenter
Planning Accounting for Power Consumption
Profiles
Sergio Nesmachnow
1
,CristianPerfumo
2
,andInigo Goiri
3
1
Universidad de la Republica, Montevideo, Uruguay
sergion@fing.edu.uy
2
CSIRO Energy, Newcastle, Australia
cristian.perfumo@csiro.au
3
Rutgers University, New Jersey, USA
goiri@cs.rutgers.edu
Abstract.
Energy eciency is one of the major concerns when operat-
ing datacenters nowadays, as the large amount of energy used by parallel
computing infrastructures impacts on both the energy cost and the elec-
tricity grid. Power consumption can be lowered by dynamically adjusting
the power demand of datacenters, but conflicting objectives such as tem-
perature and quality of service must be taken into account. This paper
proposes a multiobjective evolutionary approach to solve the energy-
aware scheduling problem in datacenters, regarding power consumption,
temperature, and quality of service when controlling servers and cool-
ing infrastructures. Accurate results are reported for both best solutions
regarding each of the problem objectives and best trade-off solutions.
1 Introduction
Nowadays, cluster/grid/cloud computing datacenters host powerful high per-
formance computing (HPC) resources having large and increasing energy de-
mands [28]. Energy eciency is a major concern in datacenter operation, as
recent surveys point out that datacenters account for about 1.5% of the total
energy usage in the world [14]. Thus, power consumption in datacenters raises
many important environmental and economic issues. Owners and operators are
highly interested in energy-ecient datacenters that applies intelligent planning
to reduce and adjust power consumption, and integrate renewable generation [8].
Within the research community, there is growing interest in energy-aware
planning of HPC infrastructures, including dynamic operation control [8][21][27],
and energy-aware scheduling and planning [28][1].
Dynamic control approaches are especially useful when considering the cost re-
ductions that the operator can achieve by shifting operation to periods of cheaper
energy prices, and the possible utilization of (intermittent) renewable solar and
wind energy. Energy-aware datacenters can also participate as providers of ancil-
lary services in the electricity market. Ancillary services are used to compensate
