Computational & Technology Resources
an online resource for computational,
engineering & technology publications
PROCEEDINGS OF THE FOURTEENTH INTERNATIONAL CONFERENCE ON CIVIL, STRUCTURAL AND ENVIRONMENTAL ENGINEERING COMPUTING
Edited by: B.H.V. Topping and P. Iványi
Rare Event Sampling using the Science Experimental Grid Laboratory
Y. Dorozhko1, K. Kratzer2, Y. Yudin1, A. Arnold2, C.W. Glass1 and M. Resch1
1High Performance Computing Center Stuttgart (HLRS)
Y. Dorozhko, K. Kratzer, Y. Yudin, A. Arnold, C.W. Glass, M. Resch, "Rare Event Sampling using the Science Experimental Grid Laboratory", in B.H.V. Topping, P. Iványi, (Editors), "Proceedings of the Fourteenth International Conference on Civil, Structural and Environmental Engineering Computing", Civil-Comp Press, Stirlingshire, UK, Paper 207, 2013. doi:10.4203/ccp.102.207
Keywords: SEGL, GriCoL, workflow, workflow management, forward flux sampling, FFS, rare event sampling, HPC, control flow, data flow, simulation technologies, workflow languages.
Rare events are difficult to simulate because of the long waiting times between their occurrences. This can be overcome by rare event sampling techniques, which allow one to push the system towards the interesting event without biasing the system. However, the implementation of these techniques, that are added on top of a conventional simulation tool, on high performance computers in parallel is a difficult task. In this paper, we describe how one of these sampling techniques, namely forward flux sampling (FFS), can be implemented as a workflow application in the context of the Science Experimental Grid Laboratory (SEGL). SEGL handles the workflow and dataflow of the simulation and matches the requirements for FFS by drawing a certain number of random datasets from a given dataspace. We demonstrate the usability of this approach on an example simulation problem from physics. FFS can easily be used together with any simulation software which can be described in the context of the SEGL framework, without the need for implementing the method by the researcher, making rare event simulations easily available. Beyond that, trajectory fragments can be calculated in parallel and on high performance computing hardware by design.
purchase the full-text of this paper (price £20)