Information Technology Reference
In-Depth Information
Where,
represents the average time a physical machine requires to finalize a
CPU intensive workload in the absent of virtual machines and processes unrelated to
those normally executed by the host operating system. On the other hand,
represents the average time the same physical machine requires to conclude the same
CPU intensive workload in presence of as many virtual machines as set in the test
scenario.
4
Results and Discussion
The figures presented in this section summarize the results for the aforementioned
tests scenarios. To ease interpretations, the Y-axis shows the intrusiveness percen-
tage, while the X-axis shows the amount of processes being executed by the end-user
during the test. It is important to note that the intrusiveness percentage is calculated as
presented in (1). Besides, each line represents the number of virtual machines simul-
taneously executed on the same desktop. Finally, acronyms are used to shorten words
as follows: Virtual Machines - VMs, VMware Workstation - VMware, and Virtual-
Box - Vbox.
First of all, as shown in Figure 1, the intrusiveness percentage measured over a
desktop based on the Intel Core 2 Duo Legacy architecture is below 4%. Similar re-
sults were obtained on the AMD Athlon architecture as depicted on Figure 2. Since
virtual machines are executed as low-priority processes, the host operating system
penalizes release and allocation of computing resources to perform them in presence
of processes executed in normal or any higher priority. Indeed, processes executed by
end-users by default are set with normal priority thus guarantying negligible intru-
siveness. Such research findings corroborate the results presented in [3] and [5].
Furthermore, these outcomes reveal that the opportunistic use of desktops based on
Intel Core 2 Duo Legacy and AMD Athlon processors can be considered as non-
intrusive. This is probably a consequence of the absent of dynamic performance and
energy-efficient technologies on both processor architectures.
Secondly, Figure 3 shows the measurements obtained on desktops which processor
architectures range from 1
st
, 2
nd
, and 4
th
generations of Intel Core processor. The line
which marker symbol is a square represents a scenario where simultaneous CPU in-
tensive workloads are executed on the same physical processor in the absent of virtual
machines. The results show that individual completion time of a task executed by an
end-user increases proportionally to the number of simultaneous tasks executed on the
same desktop, even when it is less than or equal to the number of CPU threads. Such
findings can be explained by the introduction of a selection of technologies aimed to
increase performance and energy-efficiency on recent processor architectures. As
detailed thereafter, one of these features is Intel Turbo Boost Technology [6]. This
technology automatically allows CPU cores to run faster than the base operating fre-
quency when the processor is working below rated power, temperature, and current
specification limits [6]. Therefore, Turbo Boost dynamically controls the CPU clock
frequency to be increased in presence of CPU lightweight workloads (single-threaded
or multi-threaded) and to be nominal in presence of CPU intensive workloads.