Gridbus Workflow Enactment Engine - Grid Computing: Infrastructure, Service, and Applications - page 136

TABLE 5.3
Resource Attributes
Node/Details

CPU (type/#/GHz) Middleware

manjra.cs.mu.oz.au 4/Intel Xeon/2.00 GHz SSH/GT2

vgtest.vpac.org 1/Intel Xeon/3.20 GHz SSH/GT4

ngdev.vpac.org 1/Intel Xeon/3.20 GHz SSH/GT4

brecca-1.vpac.org Intel Xeon/4/2.80 GHz SSH

brecca-2.vpac.org 4/Intel Xeon/2.80 GHz SSH

karwendel.dps.uibk.ac.at 2/AMD Opteron 880/2.39 GHz SSH/SGE

uuuu.maekawa.is.uec.ac.jp 1/Intel Xeon/2.80 GHz SSH/GT4

Walkure.maekawa.is.uec.ac.jp 1/Intel Xeon/2.80 GHz SSH/GT4

executing procedure slicer is installed on other sites. The Convert applica-
tion required to execute procedure convert is available only on the site of
the University of Melbourne. Table 5.3 shows the processor capability and
supporting middleware of each resource.
In the i rst experiment, the impact of the number of grid sites is investi-
gated for the various numbers of subjects. Figure 5.17 shows the total exe-
cution times using 1-5 grid sites for generating an atlas of 25, 50, and 100
subjects. The size of the image i le associated with each subject is around
16-22 MB. We can see that the total execution time increases as the number
of subjects increases. Additionally, the larger the number of grid sites, the
faster the execution time achieved. For example, the total execution time
of generating an atlas of 100 subjects using one grid node is 95 minutes;

100

25 Subjects
50 Subjects
100 Subjects

90

80

70

60

50

40

30

20

10

1 2 3 4 5

Number of grid sites

FIGURE 5.17 Total execution times of processing 25, 50, and 100 subjects over various
Grid sites.