A Predictive Approach for Selecting Suitable Computing Nodes in Grid Environment by Using Data Mining Technique - Advances in Computational Science and Engineering - page 201

are used to evaluate performance of this approach. The hardware information has
been described in Table 2. These nodes communicate with server via internet. Then,
the MA application is installed on nodes and broker provider application is installed
on the server computer. We divide 24 hour in to following parts:

7-9 9-12 12-15 15-17 17-20 20-22 22-24 24-2 2-7

Table 1. Assign a weight for each attribute

Name of
attributes

Weight for High
reliability

Weight for
Real-time

Weight for
Normal jobs

WH 1 =1 WR 1 =0.7 WN 1 =0.8

A 1 →

WH 2 =0.4 WR 2 =1 WN 2 =0.8

A 2 →

A 3 → WH 3 =0.7 WR 3 =0.6 WN 3 =0.6

A 4 → WH 4 =0.9 WR 4 =0.4 WN 4 =0.6

WH 5 =0.5 WR 5 =0.2 WN 5 =0.4

A 5 →

WH 6 =0.3 WR 6 =0.1 WN 6 =0.2

A 6 →

A 7 → WH 7 =0.5 WH 7 =0.9 WH 7 =0.5

A 8 → WH 8 =0.4 WH 8 =0.8 WH 8 =0.5

WH 9 =0.1 WH 9 =0.2 WH 9 =0.1

A 9 →

Table 2. Hardware information

Name Type of hardware

Node 1 Pentium4(Cache1MB),CPU2.2(INTEL), RAM 256

Node 2 Pentium(Cache2MB),CPU2.4(INTEL), RAM 512

Node 3 INTEL Pentium,CPU3.0(GLI),2, RAM 1G

Node 4 Intel(R) Core(TM)2 Duo CPU 2.16GHz RAM(3.49 GB)

Node 5 Intel(R) Core(TM)2 Duo CPU 2.16GHz RAM(3.49 GB)

Node 6 Intel(R) Core(TM)2 Duo CPU 2.16GHz RAM(3.49 GB)

Node 7 Intel(R) Core(TM)2 Duo CPU 2.16GHz RAM(2.9 GB)

HP ProLiant ML370 G4 High Performance - Intel Xeon 3.4 GHz
(2 processors)L2 cache(RAM 8G)

Node 8

Pentium4(Cache2MB),CPU3.0(INTEL), RAM 1G

server

In the first six days we have used MA Application but we didn't use the result of MA
in our broker application. Moreover we always have sent a job for all available nodes.
After that, we have activated broker provider to select only suitable nodes. Therefore,
in seventh day, we have taken the listed results in Table 3 from available nodes in the
morning in order to execution a high reliability job with size 4.47 MB and execution
time almost 18 minutes.
As you see, all eight nodes are accessible in this moment. The Table shows us, in
A2 column, Node8 is the best and Node1 is worst (in fuzzy range). When these results
have been delivered to server, broker provider on server side has selected Node4 for