Information Technology Reference
In-Depth Information
A
B
5%
5%
4%
4%
3%
3%
2%
2%
1%
1%
0%
0%
1
2
1
2
C
D
5%
5%
4%
4%
3%
3%
2%
2%
1%
1%
0%
0%
1
2
1
2
User processes
User processes
Fig. 1.
Measurements obtained on Intel Core 2 Duo processors. A and B were obtained on Intel
Core 2 Duo e7600 and C and D on Intel Core 2 Duo e6300. The results shown in the first col-
umn (to the left) were obtained by using VBox. The results shown in the second column were
obtained with VMware.
Under certain configurations and workloads, Turbo Boost technology can be used
to enable higher performance through the availability of increased core frequency [6].
Despite this fact, it also can affect the intrusiveness that end-users are able to perce-
ive when using Desktop Grids based on virtualization that run simultaneously to their
tasks. That is, when the processor is being opportunistically used by a Desktop Grid
(e.g. low-priority processes), it thermally behaves likewise being executing processes
of an end-user in normal priority. Under such circumstances, Turbo Boost technology
dynamically decreases the clock frequency thus affecting the overall performance of
the processor. As a result, this feature and other similar technologies severely impact
on the performance perceived by end-users that are donating their idle computing
resources to Desktop Grids. In fact, as can been seen on Figure 3 A and B measure-
ments show that according to the number of virtual machines opportunistically ex-
ecuted on a desktop equipped with a processor Intel Core i7-4770, the intrusiveness
percentage varies up to 60%. Furthermore, Figure 3 E and F show the intrusiveness
percentage measured over a desktop equipped with a processor Intel Core i5-660. In
this test, intrusiveness percentage is approximately 100% in the worst case. That is,
