Information Technology Reference
In-Depth Information
ACKNOWLEDGMENT
The potential benefits of the proposed algorithms
are demonstrated by performing simulations.
The simulations showed that the more the
scheduler knows about the Grid environment
and the behaviour of the job the better schedul-
ing decisions can be made and the earlier the job
completes.
This work was supported by IKTA 64/2003, OTKA
T037742, GVOP-3.3.3-2004-07-0005/3.0 ELTE
IKKK, and the Bolyai Research Fellowship.
REFERENCES
Bell, W. H., Cameron, D. G., Carvajal-Schiaffino,
R., Millar, A. P., Stockinger, K., & Zini, F. (2003).
Evaluation of an economy-based file replication
strategy for a data Grid
. In International Work-
shop on Agent based Cluster and Grid Computing
at CCGrid 2003. Tokyo, Japan: IEEE Computer
Society Press.
FUTURE WORK
The proposed scheduling algorithms disregard the
overall Grid performance and solely optimize for
the finishing time of the current job. However, the
network characteristics should also be considered,
otherwise the network capacity can become a
major bottleneck which may lead to performance
degradation. Therefore we are planning to improve
the scheduling strategies to consider the global
performance of the Grid.
According to our model the job is a single
process application running on a single processor.
We would like to relax this limitation and extend
the job behaviour description by including com-
munication patterns for applications composed
of parallel processes (e.g. PVM and MPI tasks).
Accordingly we also intend to alter the scheduling
(and estimation) algorithms to take the commu-
nication patterns into account.
The primary focus of our current work is data
intensive applications and data Grids. We would
like to generalize our approach and enable the
scheduler to make efficient decisions in such cases
when file access does not determine the execu-
tion time of the job significantly. The generalized
approach should identify those operations which
substantially influence the performance of the
job. The job behaviour description and scheduling
strategies should also be generalized to include
and consider the relevant operations.
Casanova, H., Obertelli, G., Berman, F., & Wolski,
R. (2000). The AppLeS parameter sweep template:
User-level middleware for the Grid.
Proceedings
of Supercomputing
,
00
, 75-76.
Chervenak, A., Deelman, E., Livny, M., Su, M.-
H., Schuler, R., Bharathi, S., et al. (September
2007). Data placement for scientific applications
in distributed environments.
Proceedings of the
8th IEEE/ACM International Conference on Grid
Computing (Grid2007)
.
Condor Project
. (n.d.). Retrieved from http://
www.cs.wisc.edu /condor/
Foster, I. (July 1998).
The Grid: Blueprint for a new
computing infrastructure.
Morgan-Kaufmann.
Gao, Y., Rong, H., & Huang, J. Z. (2005). Adaptive
grid job scheduling with genetic algorithms.
Fu-
ture Generation Computer Systems
,
21
, 151-161.
doi:10.1016/j.future.2004.09.033
Globus Toolkit
. (n.d.). Retrieved from http://www.
globus.org/toolkit
Job Description Language Attributes
. (n.d.). Re-
trieved from http://auger.jlab.org/jdl /PPDG_JDL.
htm
Search WWH ::
Custom Search