Information Technology Reference
In-Depth Information
Table 1. Parameters for simulations
Parameter
Value Range
No of nodes
fixed to 16
// fig. 1a to 1g
No of tasks
8 to 100
// fig. 1a to 1g
No of tasks
fixed to 40
// fig. 2a to 2g
No of nodes
4 to 24
// fig. 2a to 2g
Speed of the processing nodes (SP)
1, 2, 5, 10 (MIPS)
Security level of the processing node (SL)
4 / 6 / 9
No. of tasks
Up to 100
Size of tasks
10 to 100 (MB)
Security level demand of the grid task (SD)
4 / 6 / 9
OBSERVATIONS AND CONCLUSION
computational and security heterogeneity of
grid resources and grid tasks. Experiments are
conducted on simulated grid environments in
two sets. In the first set, the number of nodes is
fixed to 16 and the number of tasks varies from
8 to 100. In the second set number of tasks are
fixed to 100 and number of nodes varies from 4
to 24. For all possible situations, experimental
data is generated and the behavior of MinMin and
SPMinMin for makespan and site utilization are
studied and compared. The following observations
are derived from the experiments:
Whenever heavy tasks demanded more security
the proposed model always behaved much better
than MinMin as shown in Figure 1a, 1d and 2a,
2d. The more the percentage of such tasks the
more will be the improvement in performance
of SPMinMin over MinMin. When there is no
dependency between length of tasks and the se-
curity, even then SPMinMin outperforms MinMin
as depicted from plots 3c, 3f and 4c, 4f. Similar
results are obtained for both the sets; varying job
size or varying grid size. Thus for any grid size or
job size SPMinMin outperforms MinMin when-
ever heavy tasks are more security demanding.
In a grid situation when heavy tasks require
more security and higher speed nodes are more
secured, there is a significant improvement in the
performance of SPMinMin over MinMin and this
comes out to be the most favorable situation for
The present work proposes a security aware
scheduling model for computational grid as an
extension of MinMin model. It also compares the
widely used MinMin algorithm with the proposed
SPMinMin for performance metrics like makes-
pan and site utilization. SPMinMin is a modified
MinMin and it shows remarkable improvement
where security demanded by the task cannot be
compromised. According to MinMin, shortest
task will be scheduled at priority on fastest node.
This makes lighter and low security demanding
tasks to be unnecessarily scheduled on faster high
security nodes. This affects negatively on the site
utilization and the entire jobs' (multiple tasks)
makespan. To overcome this shortcoming, the
Min-Min (secured) algorithm has been modified
with the security demand of the task as a guiding
factor to scheduling decisions. The modified Se-
curity Prioritized MinMin (SPMinMin) allocates
highest security demanding tasks first on the faster
resources. The highest security requiring tasks
are then scheduled according to MinMin. Thus
it never compromises the benefits of MinMin
but very simply modifies it to work efficiently
in a secured environment. Experimental results
confirm our study.
The article has elaborated various possible
situations in a grid environment based on the
Search WWH ::
Custom Search