A Predictive Approach for Selecting Suitable Computing Nodes in Grid Environment by Using Data Mining Technique - Advances in Computational Science and Engineering - page 200

Node1
Log file

Node2
Log file

Node3
Log file

Miner-App Miner-App

Miner-App

….

Br o k er
Layer

Request broker (discovery)

Re so u r c e
monitoring

TXD

Su b m i ss i o n

Global job
queue Resource select or

Fuzzy DC execut er
Selecting nodes for job

Job
submission

GT m i d d l e w ar e
Layer

Job scheduler

MDS

Fig. 2. General architecture for proposed approach

As you see, The A 5 shows the ratio of successful jobs that have similar size with the
desired job to all successful finished jobs. A 7 shows the CPU power and A 8 shows the
measure of system Idle in fuzzy range.
For the nonce, these nine attributes will be evaluated in fuzzy behavior. We must
mention that based on the type of jobs, they will take a weight. This weight has been
allocated based on empiric and the effect of each attributed in classification by DT.
These weights are representing in Table 1. Now, to find a node with very high reliabil-
ity rather than other nodes, we should compute the following computation (formula 8)
for each node and then we will select that node with maximum Value according to
formula 9. We will have this similar method for other type of jobs.

9

(8)

∑
=

Val = i ( WH * A )

node i i

i

1

Selected _ node = a node with Max(val )

(9)

node

i

5 Experimental Results and Discussion

We have designed two applications for our approach. The first application is executed
on nodes (MA). In our experiments, eight resource computing nodes and one server