Information Technology Reference
In-Depth Information
end-users. As a result, we demonstrate that intrusiveness caused by Desktop Grids
based on virtualization varies according to dynamic performance and energy-efficient
technologies enabled on the supporting processor. To the best of our knowledge, our
work is the first to evaluate how dynamic performance and energy-efficient technolo-
gies, as features introduced in modern processor architectures, affect the intrusiveness
that Desktop Grids based on virtualization generate on desktops. Furthermore, a first
set of strategies are recommended to implementations of Desktop Grid systems in
order to reduce degradation on the performance perceived by end-users when donat-
ing their resources to Desktop Grid Computing.
3
Methodology
The aim of this research is to show implications of technologies incorporated on re-
cent processor architectures in relation to intrusiveness produced by Desktop Grids.
Specifically, our hypothesis is that dynamic performance and energy-efficient tech-
nologies directly affect the intrusiveness that Desktop Grids supported by virtualiza-
tion generate when running over desktops. We clarify that our research findings may
not be limited to Desktop Grids based on virtualization, but they may be generalized
to every type of Desktop Grid; however, such generalization cannot be confirmed
from the limited data obtained in this research. In addition, our research outcomes are
not limited to the generations of processors studied, but they can be obtained in simi-
lar processor architectures, including servers. Nevertheless, the scope of this paper
considers a specific selection of processors and a unique Desktop Grid system sup-
ported by virtualization technologies. In consequence, a set of test scenarios based on
virtualization are proposed and performed over a variety of desktops. Finally, we
clarify the test scenarios were designed in order to exclusively evaluate intrusiveness
in terms of the performance delivered by the processor. To validate our hypotheses,
the following methodology was developed.
3.1
Experimental Setup
The tests were executed over different generations of off-the-shelf desktop comput-
ers from laboratories open to employees, professors, and students on our university
campus. Moreover, such computers have been opportunistically exploited in order to
execute virtual machines for multiple e-Science projects [11]. The variety of such
desktops was constrained by the inventory of hardware on our campus. This invento-
ry is updated annually, hence limiting the amount of processor architectures to be
studied. All of these desktops had Windows 7 as host operating system. This selec-
tion is justified by the fact that Windows is the dominating operating system for
Desktop Grids [1]. The hardware configurations used in the experiments are shown
in Table 1.
