CYPRESS: Combining Static and Dynamic Analysis for Top-Down Communication Trace Compression

Jidong Zhai, Jianfei Hu, Xiongchao Tang, Xiaosong Ma, Wenguang Chen

Research output: Chapter in Book/Report/Conference proceedingConference contribution

12 Citations (Scopus)

Abstract

Communication traces are increasingly important, both for parallel applications' performance analysis/optimization, and for designing next-generation HPC systems. Meanwhile, the problem size and the execution scale on supercomputers keep growing, producing prohibitive volume of communication traces. To reduce the size of communication traces, existing dynamic compression methods introduce large compression overhead with the job scale. We propose a hybrid static-dynamic method that leverages information acquired from static analysis to facilitate more effective and efficient dynamic trace compression. Our proposed scheme, Cypress, extracts a program communication structure tree at compile time using inter-procedural analysis. This tree naturally contains crucial iterative computing features such as the loop structure, allowing subsequent runtime compression to "fill in", in a "top-down" manner, event details into the known communication template. Results show that Cypress reduces intra-process and inter-process compression overhead up to 5× and 9× respectively over state-of-the-art dynamic methods, while only introducing very low compiling overhead.

Original languageEnglish
Title of host publicationInternational Conference for High Performance Computing, Networking, Storage and Analysis, SC
PublisherIEEE Computer Society
Pages143-153
Number of pages11
Volume2015-January
EditionJanuary
DOIs
Publication statusPublished - 16 Jan 2015
EventInternational Conference for High Performance Computing, Networking, Storage and Analysis, SC 2014 - New Orleans, United States
Duration: 16 Nov 201421 Nov 2014

Other

OtherInternational Conference for High Performance Computing, Networking, Storage and Analysis, SC 2014
CountryUnited States
CityNew Orleans
Period16/11/1421/11/14

Fingerprint

Static analysis
Dynamic analysis
Compaction
Communication
Supercomputers

Keywords

  • High Performance Computing
  • Message Passing
  • Performance Analysis
  • Trace Compression

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Computer Science Applications
  • Hardware and Architecture
  • Software

Cite this

Zhai, J., Hu, J., Tang, X., Ma, X., & Chen, W. (2015). CYPRESS: Combining Static and Dynamic Analysis for Top-Down Communication Trace Compression. In International Conference for High Performance Computing, Networking, Storage and Analysis, SC (January ed., Vol. 2015-January, pp. 143-153). [7012999] IEEE Computer Society. https://doi.org/10.1109/SC.2014.17

CYPRESS : Combining Static and Dynamic Analysis for Top-Down Communication Trace Compression. / Zhai, Jidong; Hu, Jianfei; Tang, Xiongchao; Ma, Xiaosong; Chen, Wenguang.

International Conference for High Performance Computing, Networking, Storage and Analysis, SC. Vol. 2015-January January. ed. IEEE Computer Society, 2015. p. 143-153 7012999.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Zhai, J, Hu, J, Tang, X, Ma, X & Chen, W 2015, CYPRESS: Combining Static and Dynamic Analysis for Top-Down Communication Trace Compression. in International Conference for High Performance Computing, Networking, Storage and Analysis, SC. January edn, vol. 2015-January, 7012999, IEEE Computer Society, pp. 143-153, International Conference for High Performance Computing, Networking, Storage and Analysis, SC 2014, New Orleans, United States, 16/11/14. https://doi.org/10.1109/SC.2014.17
Zhai J, Hu J, Tang X, Ma X, Chen W. CYPRESS: Combining Static and Dynamic Analysis for Top-Down Communication Trace Compression. In International Conference for High Performance Computing, Networking, Storage and Analysis, SC. January ed. Vol. 2015-January. IEEE Computer Society. 2015. p. 143-153. 7012999 https://doi.org/10.1109/SC.2014.17
Zhai, Jidong ; Hu, Jianfei ; Tang, Xiongchao ; Ma, Xiaosong ; Chen, Wenguang. / CYPRESS : Combining Static and Dynamic Analysis for Top-Down Communication Trace Compression. International Conference for High Performance Computing, Networking, Storage and Analysis, SC. Vol. 2015-January January. ed. IEEE Computer Society, 2015. pp. 143-153
@inproceedings{7e239fa85f4745b6932c291ee694c8d0,
title = "CYPRESS: Combining Static and Dynamic Analysis for Top-Down Communication Trace Compression",
abstract = "Communication traces are increasingly important, both for parallel applications' performance analysis/optimization, and for designing next-generation HPC systems. Meanwhile, the problem size and the execution scale on supercomputers keep growing, producing prohibitive volume of communication traces. To reduce the size of communication traces, existing dynamic compression methods introduce large compression overhead with the job scale. We propose a hybrid static-dynamic method that leverages information acquired from static analysis to facilitate more effective and efficient dynamic trace compression. Our proposed scheme, Cypress, extracts a program communication structure tree at compile time using inter-procedural analysis. This tree naturally contains crucial iterative computing features such as the loop structure, allowing subsequent runtime compression to {"}fill in{"}, in a {"}top-down{"} manner, event details into the known communication template. Results show that Cypress reduces intra-process and inter-process compression overhead up to 5× and 9× respectively over state-of-the-art dynamic methods, while only introducing very low compiling overhead.",
keywords = "High Performance Computing, Message Passing, Performance Analysis, Trace Compression",
author = "Jidong Zhai and Jianfei Hu and Xiongchao Tang and Xiaosong Ma and Wenguang Chen",
year = "2015",
month = "1",
day = "16",
doi = "10.1109/SC.2014.17",
language = "English",
volume = "2015-January",
pages = "143--153",
booktitle = "International Conference for High Performance Computing, Networking, Storage and Analysis, SC",
publisher = "IEEE Computer Society",
edition = "January",

}

TY - GEN

T1 - CYPRESS

T2 - Combining Static and Dynamic Analysis for Top-Down Communication Trace Compression

AU - Zhai, Jidong

AU - Hu, Jianfei

AU - Tang, Xiongchao

AU - Ma, Xiaosong

AU - Chen, Wenguang

PY - 2015/1/16

Y1 - 2015/1/16

N2 - Communication traces are increasingly important, both for parallel applications' performance analysis/optimization, and for designing next-generation HPC systems. Meanwhile, the problem size and the execution scale on supercomputers keep growing, producing prohibitive volume of communication traces. To reduce the size of communication traces, existing dynamic compression methods introduce large compression overhead with the job scale. We propose a hybrid static-dynamic method that leverages information acquired from static analysis to facilitate more effective and efficient dynamic trace compression. Our proposed scheme, Cypress, extracts a program communication structure tree at compile time using inter-procedural analysis. This tree naturally contains crucial iterative computing features such as the loop structure, allowing subsequent runtime compression to "fill in", in a "top-down" manner, event details into the known communication template. Results show that Cypress reduces intra-process and inter-process compression overhead up to 5× and 9× respectively over state-of-the-art dynamic methods, while only introducing very low compiling overhead.

AB - Communication traces are increasingly important, both for parallel applications' performance analysis/optimization, and for designing next-generation HPC systems. Meanwhile, the problem size and the execution scale on supercomputers keep growing, producing prohibitive volume of communication traces. To reduce the size of communication traces, existing dynamic compression methods introduce large compression overhead with the job scale. We propose a hybrid static-dynamic method that leverages information acquired from static analysis to facilitate more effective and efficient dynamic trace compression. Our proposed scheme, Cypress, extracts a program communication structure tree at compile time using inter-procedural analysis. This tree naturally contains crucial iterative computing features such as the loop structure, allowing subsequent runtime compression to "fill in", in a "top-down" manner, event details into the known communication template. Results show that Cypress reduces intra-process and inter-process compression overhead up to 5× and 9× respectively over state-of-the-art dynamic methods, while only introducing very low compiling overhead.

KW - High Performance Computing

KW - Message Passing

KW - Performance Analysis

KW - Trace Compression

UR - http://www.scopus.com/inward/record.url?scp=84936943828&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84936943828&partnerID=8YFLogxK

U2 - 10.1109/SC.2014.17

DO - 10.1109/SC.2014.17

M3 - Conference contribution

AN - SCOPUS:84936943828

VL - 2015-January

SP - 143

EP - 153

BT - International Conference for High Performance Computing, Networking, Storage and Analysis, SC

PB - IEEE Computer Society

ER -