Neighborhood communication paradigm to increase scalability in large-scale dynamic scientific applications

Handle URI:
http://hdl.handle.net/10754/598964
Title:
Neighborhood communication paradigm to increase scalability in large-scale dynamic scientific applications
Authors:
Ovcharenko, Aleksandr; Ibanez, Daniel; Delalondre, Fabien; Sahni, Onkar; Jansen, Kenneth E.; Carothers, Christopher D.; Shephard, Mark S.
Abstract:
This paper introduces a general-purpose communication package built on top of MPI which is aimed at improving inter-processor communications independently of the supercomputer architecture being considered. The package is developed to support parallel applications that rely on computation characterized by large number of messages of various sizes, often small, that are focused within processor neighborhoods. In some cases, such as solvers having static mesh partitions, the number and size of messages are known a priori. However, in other cases such as mesh adaptation, the messages evolve and vary in number and size and include the dynamic movement of partition objects. The current package provides a utility for dynamic applications based on two key attributes that are: (i) explicit consideration of the neighborhood communication pattern to avoid many-to-many calls and also to reduce the number of collective calls to a minimum, and (ii) use of non-blocking MPI functions along with message packing to manage message flow control and reduce the number and time of communication calls. The test application demonstrated is parallel unstructured mesh adaptation. Results on IBM Blue Gene/P and Cray XE6 computers show that the use of neighborhood-based communication control leads to scalable results when executing generally imbalanced mesh adaptation runs. © 2011 Elsevier B.V. All rights reserved.
Citation:
Ovcharenko A, Ibanez D, Delalondre F, Sahni O, Jansen KE, et al. (2012) Neighborhood communication paradigm to increase scalability in large-scale dynamic scientific applications. Parallel Computing 38: 140–156. Available: http://dx.doi.org/10.1016/j.parco.2011.10.013.
Publisher:
Elsevier BV
Journal:
Parallel Computing
Issue Date:
Mar-2012
DOI:
10.1016/j.parco.2011.10.013
Type:
Article
ISSN:
0167-8191
Sponsors:
This work is supported by the National Science Foundation under Grant No. 0749152, and by the US Department of Energy under DOE Grant No. DE-FC02-06ER25769. Computing support is provided by King Abdullah University of Science and Technology for granting access to the Blue Gene/P and by National Energy Research Scientific Computing Center for granting access to the Cray XE6 supercomputers.The authors thank Professor William D. Gropp for the valuable comments and clarifications towards the improvement of this study.
Appears in Collections:
Publications Acknowledging KAUST Support

Full metadata record

DC FieldValue Language
dc.contributor.authorOvcharenko, Aleksandren
dc.contributor.authorIbanez, Danielen
dc.contributor.authorDelalondre, Fabienen
dc.contributor.authorSahni, Onkaren
dc.contributor.authorJansen, Kenneth E.en
dc.contributor.authorCarothers, Christopher D.en
dc.contributor.authorShephard, Mark S.en
dc.date.accessioned2016-02-25T13:44:34Zen
dc.date.available2016-02-25T13:44:34Zen
dc.date.issued2012-03en
dc.identifier.citationOvcharenko A, Ibanez D, Delalondre F, Sahni O, Jansen KE, et al. (2012) Neighborhood communication paradigm to increase scalability in large-scale dynamic scientific applications. Parallel Computing 38: 140–156. Available: http://dx.doi.org/10.1016/j.parco.2011.10.013.en
dc.identifier.issn0167-8191en
dc.identifier.doi10.1016/j.parco.2011.10.013en
dc.identifier.urihttp://hdl.handle.net/10754/598964en
dc.description.abstractThis paper introduces a general-purpose communication package built on top of MPI which is aimed at improving inter-processor communications independently of the supercomputer architecture being considered. The package is developed to support parallel applications that rely on computation characterized by large number of messages of various sizes, often small, that are focused within processor neighborhoods. In some cases, such as solvers having static mesh partitions, the number and size of messages are known a priori. However, in other cases such as mesh adaptation, the messages evolve and vary in number and size and include the dynamic movement of partition objects. The current package provides a utility for dynamic applications based on two key attributes that are: (i) explicit consideration of the neighborhood communication pattern to avoid many-to-many calls and also to reduce the number of collective calls to a minimum, and (ii) use of non-blocking MPI functions along with message packing to manage message flow control and reduce the number and time of communication calls. The test application demonstrated is parallel unstructured mesh adaptation. Results on IBM Blue Gene/P and Cray XE6 computers show that the use of neighborhood-based communication control leads to scalable results when executing generally imbalanced mesh adaptation runs. © 2011 Elsevier B.V. All rights reserved.en
dc.description.sponsorshipThis work is supported by the National Science Foundation under Grant No. 0749152, and by the US Department of Energy under DOE Grant No. DE-FC02-06ER25769. Computing support is provided by King Abdullah University of Science and Technology for granting access to the Blue Gene/P and by National Energy Research Scientific Computing Center for granting access to the Cray XE6 supercomputers.The authors thank Professor William D. Gropp for the valuable comments and clarifications towards the improvement of this study.en
dc.publisherElsevier BVen
dc.subjectAsynchronous communicationen
dc.subjectDynamic data migrationen
dc.subjectMPIen
dc.subjectOverlapping communication and computationen
dc.subjectParallel algorithmsen
dc.titleNeighborhood communication paradigm to increase scalability in large-scale dynamic scientific applicationsen
dc.typeArticleen
dc.identifier.journalParallel Computingen
dc.contributor.institutionRensselaer Polytechnic Institute, Troy, United Statesen
dc.contributor.institutionUniversity of Colorado at Boulder, Boulder, United Statesen
All Items in KAUST are protected by copyright, with all rights reserved, unless otherwise indicated.