Information Technology Reference
In-Depth Information
sequential application, the processing times are conditioned by the time required
to complete the most CPU demanding task.
No optimization is applied for grouping tasks as the processing times for
each task is unknown before hand. As mentioned earlier in this section, groups
are created using uniform distributions and random selection. This method of
selection explains why a group of 6 tasks (group #3) is processed faster than all
groups of 4 tasks. Even though a subset of results is obtained earlier, the total
processing time still remains comparable.
As the main goal of this work is to reduce the time needed to process the
whole set of instances, the results obtained using groups of different sizes are
valid. The speedup factor obtained is approximately 4.
6 Conclusions and Future Work
This article presented an experimental analysis of applying a grid/cloud ap-
proach for the execution of different scenarios for fluorescence analysis. Results
for sequential and parallel execution of MCell and FERNet were reported.
The experimental evaluation of the proposed distributed computing approach
was carried out using a volunteer federation of OurGrid sites distributed in four
countries in Latin America. The eciency analysis demonstrates that the use
of a volunteer grid/cloud infrastructure is an effective method for reducing the
overall execution time of simulations. Overall, execution time reductions up to
about
70-75%
were obtained when solving different scenarios for the considered
problem. These results suggest that the volunteer computing paradigm suits well
for executing simulations of complex biological phenomena.
The main lines for future work include increasing the number of models and
scenarios to be simulated and performing more realistic experiments including
more biologically-relevant parameters of the models. Additional studies regard-
ing the composition of groups of tasks should also be performed, as well as further
extending the scalability analysis when using distributed infrastructures for solve
very complex problems. We are also working on executing MCell and FERnet on
federations of volunteers OpenNebula sites running Hadoop Framework. Besides
that, tests on Microsoft Azure Cloud Platform are also being carried out.
Acknowledgements.
This work was supported by grants from Universidad de
Buenos Aires (UBACYT 2011-2014 GC 20020100100889) and CONICET (PIP
GI 11220110100379), Argentina. The work of M. Da Silva and S. Nesmachnow
is partly funded by ANII and PEDECIBA, Uruguay. M. Geier and J. Angiolini
have scholarship from CONICET, Argentina.
References
1. Anderson, D.: BOINC: A system for public-resource computing and storage. In:
5th Int. Workshop on Grid Computing, Pittsburgh, USA, pp. 4-10 (2004)
2. Anderson, D., Fedak, G.: The computational and storage potential of volunteer com-
puting. In: 6th Int. Symp. on Cluster Computing and the Grid, pp. 73-80 (2006)
