Information Technology Reference
In-Depth Information
Figure 1. Turnaround Time v/s Number of Failed Nodes
is too low, the result could not be accommodated
along with other results shown in Figure 2. Figure
1 and Figure 2 corresponds to the experimental
results without using any replication.
It is observed that the turnaround time keeps
on increasing with increase in the number of fail-
ing nodes. In addition, replication of modules
increases the reliability of job execution. This
reliability is minimum when maximum number
of nodes has failed and increases with reducing
number of failing nodes. Thus, in spite of the node
failures jobs gets executed with an increased
turnaround time adhering to the purpose of the
RBS.
Figure 3 and Figure 4 present the effect on
grid reliability due to replication. It presents a
comparison of the turnaround time and reliabil-
ity obtained with and without node replication
keeping the same grid environment and jobs.
The experiments were set to have no node failure
when no replication is there. The turnaround time
and reliability values are observed. Also, experi-
ment incorporated node failure feature along with
module replication and the experiments were run
again in the same grid environment for the same
job. The turnaround time and reliability values
with replication incorporated as reported here
correspond to the ones with minimum nodes failed
in each experiment.
It is evident from Figure 3 that the turnaround
time increases with the introduction of replication
owing to the cost of node failures resulting in
execution of the job from the replicas. Thus the
job gets executed though with some inflated
turnaround time. Since, the job is getting execut-
ed due to the presence of replicas; it results in an
increased reliability for the job as conspicuous in
Figure 4. Thus the presence of replica ensures an
increased reliability for the job execution. Same
pattern, as reported in Figure 1 to Figure 4 is
noticed in many more experiments validating the
performance of the model.
CONCLUSION
The proposed Replica Based co-scheduler (RBS)
helps in the reliable execution of the modular job
by replicating the modules allocated to the nodes
with high failure rates (sick nodes) to the ones with
a lower failure rates (healthy nodes). In place of
having full redundancy, partial redundancy has
Search WWH ::




Custom Search