Modern data centers are continuously expanding as they
attempt to accommodate the surging scientific and enterprise
demand for computing resources. Driven by this fast-paced
demand, the cloud computing paradigm has emerged as a
key to leverage virtualization technologies and to enable a
more efficient resource management. In this context, however,
energy consumption becomes a critical concern for large-scale
datacenters, as well as for the growing cloud infrastructures
they host. While cloud computing holds the promise to deliver
unlimited processing power on demand, its requirements for
large-scale resources and the associated costs have shifted
the research focus from optimizing performance to finding a
tradeoff between performance and energy efficiency.
Existing works focus on profiling the energy usage of
hardware components or the virtualization overhead introduced
by cloud environments, yet fail to assess how the overall cloud
energy consumption is impacted. More advanced works propose
consolidation and resource allocation strategies to build
energy-efficient platforms. Nevertheless, they do not consider
the effects of cluster customization and virtual machine (VM)
settings on the platform power profile.
The goal of this paper is to investigate such aspects
and to provide an in-depth understanding regarding energy
consumption dynamics in Infrastructure-as-a-Service (IaaS)
environments. Our main contribution is twofold. First, we
provide an evaluation of two well known IaaS cloud platforms
with respect to energy consumption. Second, we study the
impact of virtual cluster management on the cloud power
profile. We discuss the effects of executing applications in
customized virtual clusters both in terms of application performance
and energy usage. Our objective is to provide users
with a basis and a set of guidelines for making energy-aware
decisions when selecting specific cloud infrastructures and VM
configurations for typical cloud workloads. Our findings can
serve as a starting point to build higher-level services capable
of achieving overall energy reductions at the cloud level, while
delivering the same performance to the users.
The reminder of this paper is structured as follows. Section
II provides a review of the existing approaches to optimize
energy consumption in IT infrastructures. In Section III, we
introduce the cloud platforms we investigate, followed by a
detailed description of the use case applications in Section IV.
The experimental setup and used metrics are presented in Section
V. and Section VI introduces and analyzes our evaluation
results. Finally, Section VII highlights our key findings, while
Section VIII draws conclusions and directions for future work.