Information Technology Reference
In-Depth Information
the overall execution time can be considerably reduced —
up to 24 times
—and
can be applied on monolithic legacy code used in geological modeling applications
(like
sisim
) within real scenarios, making it practical to be used in industrial
workflows. It is worth to mention that this strategy is general enough to be used
in other analogous domains.
The entire system implementation exercise presented different challenges re-
lated to distributed systems that were solved using cutting edges approaches.
For example, multiple processes synchronization and node communication were
implemented with message passing; data handling, via a distributed Document-
oriented NoSQL database; and, cluster management, with a remote configuration
management tool.
The presented work was focused on a particular algorithm, but this (general)
strategy can be applied to a wide range of applications. In geosciences only, there
are many other algorithms that have the same characteristics, like being embar-
rassingly parallelizable or having heavy task workflows (Sequential Gaussian
Simulation [7], Turning Bands Methods [9], Multi-point Statistics Algorithms
[14], among others). This point is a motivation to replicate this architecture
many times. Thus, an ecient way to do this, is to build a general framework
software to easily generate parallel and distributed executions, specially focused
in scientific applications. This framework could be a helpful tool for researchers
that need to generate rapid software prototypes that includes HPC features,
allowing to integrate new and existing code with a small effort.
Acknowledgments.
The authors would like to thank the industrial supporters
of ALGES laboratory, in particular BHP Billiton, Codelco Chile and Yamana
Gold, as well as the support from the Advanced Mining Technology Center
(AMTC) and the whole ALGES team.
References
1. Gridgain = in-memory computing platform computing,
http://www.gridgain.com
2. Proactive: Open source solution for parallel, distributed, multi-core computing,
3. Remics: Reuse and migration of legacy applications to interoperable cloud services,
4. Armstrong, M.P., Marciano, R.J.: Massively parallel strategies for local spatial
interpolation. Computers & Geosciences 23(8), 859-867 (1997)
5. Bergen, A., Yazir, Y.O., Muller, H.A., Coady, Y.: RPC automation: Making legacy
code relevant. In: 2013 ICSE Workshop on Software Engineering for Adaptive and
Self-Managing Systems (SEAMS), pp. 175-180 (2013)
6. Cheng, T.: Accelerating universal Kriging interpolation algorithm using CUDA-
enabled GPU. Computers & Geosciences 54, 178-183 (2013)
7. Deutsch, C., Journel, A.: GSLIB: Geostatistical software library and users guide.
Oxford University Press, New York (1998)
8. Dworak, A., Charrue, P., Ehm, F., Sliwinski, W., Sobczak, M.: Middleware Trends
and Market Leaders 2011. In: 13th International Conference on Accelerator and
Large Experimental Physics Control Systems, p. 1334 (2011)
