Moldable Job Allocation for Handling Resource Fragmentation in Computational Grid - Cloud, Grid and High Performance Computing: Emerging Applications - page 71

Figure 7. Comparison under SDSC's SP2 workload for uniformly and normally distributed slowdown
values

slowdown value which is determined by both the
parallel program characteristics and underlying
interconnection speed. On the other hand, perfor-
mance of moldable job allocation is irrelative to
the slowdown values and the results also indicate
that moldable job allocation outperforms multi-
site co-allocation in the simulations.

when parallel jobs have the moldable property.
The effects of inexact runtime estimations on
system performance are also investigated. The
results indicate that the proposed moldable job
allocation policies are effective as well as stable
under different system configurations and can
tolerate a wide range of estimation errors.

CONCLUSION REFERENCES

Barsanti, L., & Sodan, A. (2007). Adaptive job
scheduling via predictive job resource alloca-
tion. Proceedings of the 12 th Conference on Job
Scheduling Strategies for Parallel Processing,
(pp. 115-140).

In the real world, a Grid environment is usually
heterogeneous at least for the different comput-
ing speeds at different participating sites. The
heterogeneity presents a challenge for effectively
arranging load sharing activities in a computational
Grid. This paper develops moldable job allocation
policies based on the moldable property of paral-
lel applications for heterogeneous computational
Grids. The proposed policies can be used when a
parallel job, during the scheduling activities, can-
not fit in any single site in the Grid. The proposed
policies require users to provide estimations of
job execution times upon job submission. The
policies are evaluated through a series of simu-
lations using real workload traces. The results
indicate that the moldable job allocation policies
can further improve the system performance of a
heterogeneous computational Grid significantly

Buyya, R., Abramson, D., Giddy, J., & Stockinger,
H. (2002). Economic models for resource manage-
ment and scheduling in Grid computing. Concur-
rency and Computation , 14 (13-15), 1507-1542.
doi:10.1002/cpe.690

Buyya, R., Giddy, J., & Abramson, D. (2000). An
evaluation of economy-based resource trading
and scheduling on computational power grids for
parameter sweep applications. Paper presented
at the Second Workshop on Active Middleware
Services (AMS2000), Pittsburgh, USA.