Runtime MPI collective checking with tree-based overlay networks

Tobias Hilbrich; Fabian Hänsel; Martin Schulz; Bronis R. De Supinski; Matthias S. Müller; Wolfgang E. Nagel

doi:10.1145/2488551.2488570

Runtime MPI collective checking with tree-based overlay networks

Tobias Hilbrich, Fabian Hänsel, Martin Schulz, Bronis R. De Supinski, Matthias S. Müller, Wolfgang E. Nagel

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

12 Scopus citations

Abstract

Runtime error detection tools detect many classes of MPI usage errors, including errors in collective communication calls. However, they often face scalability challenges. We present runtime checks for MPI collective operations that use a Tree-Based Overlay Network (TBON) for scalability and that provide full datatype matching. While we can use transitive correctness properties for most checks, some collective operations impose non-transitive correctness properties, e.g., MPI-Alltoallv, where we use an intralayer communication within the TBON to distribute datatype matching information. An overhead study with stress tests and two benchmark suites demonstrates applicability and scalability at 4,096, 2,048 and 16,384 processes respectively.

Original language	English
Title of host publication	Proceedings of the 20th European MPI Users' Group Meeting, EuroMPI 2013
Publisher	Association for Computing Machinery
Pages	129-134
Number of pages	6
ISBN (Print)	9788461651337
DOIs	https://doi.org/10.1145/2488551.2488570
State	Published - 2013
Externally published	Yes
Event	20th European MPI Users' Group Meeting, EuroMPI 2013 - Madrid, Spain Duration: 15 Sep 2013 → 18 Sep 2013

Publication series

Name	ACM International Conference Proceeding Series

Conference

Conference	20th European MPI Users' Group Meeting, EuroMPI 2013
Country/Territory	Spain
City	Madrid
Period	15/09/13 → 18/09/13

Keywords

Correctness
MPI collectives
Tree-based overlay networks

Access to Document

10.1145/2488551.2488570

Cite this

Hilbrich, T., Hänsel, F., Schulz, M., De Supinski, B. R., Müller, M. S., & Nagel, W. E. (2013). Runtime MPI collective checking with tree-based overlay networks. In Proceedings of the 20th European MPI Users' Group Meeting, EuroMPI 2013 (pp. 129-134). (ACM International Conference Proceeding Series). Association for Computing Machinery. https://doi.org/10.1145/2488551.2488570

@inproceedings{ea5170a7580f4918b414ba2dc6e2becf,

title = "Runtime MPI collective checking with tree-based overlay networks",

abstract = "Runtime error detection tools detect many classes of MPI usage errors, including errors in collective communication calls. However, they often face scalability challenges. We present runtime checks for MPI collective operations that use a Tree-Based Overlay Network (TBON) for scalability and that provide full datatype matching. While we can use transitive correctness properties for most checks, some collective operations impose non-transitive correctness properties, e.g., MPI-Alltoallv, where we use an intralayer communication within the TBON to distribute datatype matching information. An overhead study with stress tests and two benchmark suites demonstrates applicability and scalability at 4,096, 2,048 and 16,384 processes respectively.",

keywords = "Correctness, MPI collectives, Tree-based overlay networks",

author = "Tobias Hilbrich and Fabian H{\"a}nsel and Martin Schulz and {De Supinski}, {Bronis R.} and M{\"u}ller, {Matthias S.} and Nagel, {Wolfgang E.}",

year = "2013",

doi = "10.1145/2488551.2488570",

language = "English",

isbn = "9788461651337",

series = "ACM International Conference Proceeding Series",

publisher = "Association for Computing Machinery",

pages = "129--134",

booktitle = "Proceedings of the 20th European MPI Users' Group Meeting, EuroMPI 2013",

note = "20th European MPI Users' Group Meeting, EuroMPI 2013 ; Conference date: 15-09-2013 Through 18-09-2013",

}

Hilbrich, T, Hänsel, F, Schulz, M, De Supinski, BR, Müller, MS & Nagel, WE 2013, Runtime MPI collective checking with tree-based overlay networks. in Proceedings of the 20th European MPI Users' Group Meeting, EuroMPI 2013. ACM International Conference Proceeding Series, Association for Computing Machinery, pp. 129-134, 20th European MPI Users' Group Meeting, EuroMPI 2013, Madrid, Spain, 15/09/13. https://doi.org/10.1145/2488551.2488570

Runtime MPI collective checking with tree-based overlay networks. / Hilbrich, Tobias; Hänsel, Fabian; Schulz, Martin et al.
Proceedings of the 20th European MPI Users' Group Meeting, EuroMPI 2013. Association for Computing Machinery, 2013. p. 129-134 (ACM International Conference Proceeding Series).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Runtime MPI collective checking with tree-based overlay networks

AU - Hilbrich, Tobias

AU - Hänsel, Fabian

AU - Schulz, Martin

AU - De Supinski, Bronis R.

AU - Müller, Matthias S.

AU - Nagel, Wolfgang E.

PY - 2013

Y1 - 2013

N2 - Runtime error detection tools detect many classes of MPI usage errors, including errors in collective communication calls. However, they often face scalability challenges. We present runtime checks for MPI collective operations that use a Tree-Based Overlay Network (TBON) for scalability and that provide full datatype matching. While we can use transitive correctness properties for most checks, some collective operations impose non-transitive correctness properties, e.g., MPI-Alltoallv, where we use an intralayer communication within the TBON to distribute datatype matching information. An overhead study with stress tests and two benchmark suites demonstrates applicability and scalability at 4,096, 2,048 and 16,384 processes respectively.

AB - Runtime error detection tools detect many classes of MPI usage errors, including errors in collective communication calls. However, they often face scalability challenges. We present runtime checks for MPI collective operations that use a Tree-Based Overlay Network (TBON) for scalability and that provide full datatype matching. While we can use transitive correctness properties for most checks, some collective operations impose non-transitive correctness properties, e.g., MPI-Alltoallv, where we use an intralayer communication within the TBON to distribute datatype matching information. An overhead study with stress tests and two benchmark suites demonstrates applicability and scalability at 4,096, 2,048 and 16,384 processes respectively.

KW - Correctness

KW - MPI collectives

KW - Tree-based overlay networks

UR - http://www.scopus.com/inward/record.url?scp=84886250838&partnerID=8YFLogxK

U2 - 10.1145/2488551.2488570

DO - 10.1145/2488551.2488570

M3 - Conference contribution

AN - SCOPUS:84886250838

SN - 9788461651337

T3 - ACM International Conference Proceeding Series

SP - 129

EP - 134

BT - Proceedings of the 20th European MPI Users' Group Meeting, EuroMPI 2013

PB - Association for Computing Machinery

T2 - 20th European MPI Users' Group Meeting, EuroMPI 2013

Y2 - 15 September 2013 through 18 September 2013

ER -

Runtime MPI collective checking with tree-based overlay networks

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this