Skip to main content
SpringerLink
Log in

A transmission optimization method for MPI communications

  • Published:
The Journal of Supercomputing Aims and scope Submit manuscript

Abstract

In recent years, MPI has been widely used as a communication protocol for massively parallel computing tasks, and the performance of MPI interprocess communications has become a major constraint for large-scale scalability. By analyzing the performance characteristics of bandwidth and latency of MPI communications, a transmission optimization method for MPI communications is proposed. For the variables of transmitted data, the communication strategy of MPI is optimized according to the data size and the succession of multiple communications, and the operation of packing or unpacking is automatically selected, which makes the performance of MPI communications significantly improved. For the time-consuming parts of MPI communication in the ocean numerical model Parallel Ocean Program with this method used, at least 2.4x speedup in point-to-point communication with unpacking strategy and at least 1.7x speedup in point-to-point with packing strategy are achieved. By automating file scans and analysis, 1.6x speedup is achieved for some point-to-point communications.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8

Similar content being viewed by others

Availability of data and materials

The authors confirm that the data supporting the findings of this study is available within the article.

References

  1. Wang Y, Jiang J, Zhang J, He J, Zhang H, Chi X, Yue T (2018) An efficient parallel algorithm for the coupling of global climate models and regional climate models on a large-scale multi-core cluster. J Supercomput 74:3999–4018. https://doi.org/10.1007/s11227-018-2406-6

    Article  Google Scholar 

  2. Li H, Luan ZZ (2013) A performance tool for earth system models development. Adv Mater Res 756:3814–3820. https://doi.org/10.4028/www.scientific.net/AMR.756-759.3814

    Article  Google Scholar 

  3. Zeng Y, Wang L, Zhang J, Zhu G, Zhuang Y, Guo Q (2020) Redistributing and optimizing high-resolution ocean model pop2 to million sunway cores. In: Qiu M (ed) Algorithms and architectures for parallel processing. Springer, Cham, pp 275–289. https://doi.org/10.1007/978-3-030-60245-1_19

    Chapter  Google Scholar 

  4. Suresh KK, Ramesh B, Ghazimirsaeed SM, Bayatpour M, Hashmi J, Panda DK (2020) Performance characterization of network mechanisms for non-contiguous data transfers in mpi. In: 2020 IEEE international parallel and distributed processing symposium workshops (IPDPSW), pp 896–905. https://doi.org/10.1109/IPDPSW50202.2020.00150

  5. Castain RH, Solt D, Hursey J, Bouteiller A (2017) Pmix: Process management for exascale environments. In: Proceedings of the 24th European MPI users’ group meeting. Association for Computing Machinery, New York. https://doi.org/10.1145/3127024.3127027

  6. Zheng W, Fang J, Juan C, Wu F, Pan X, Wang H, Sun X, Yuan Y, Xie M, Huang C et al (2019) Auto-tuning mpi collective operations on large-scale parallel systems. In: 2019 IEEE 21st International Conference on High Performance Computing and Communications; IEEE 17th International Conference on Smart City; IEEE 5th International Conference on Data Science and Systems (HPCC/SmartCity/DSS). IEEE, pp 670–677. https://doi.org/10.1109/HPCC/SmartCity/DSS.2019.00101

  7. Hunold S, Carpen-Amarie A, Lübbe FD, Träff JL (2016) Pgmpi: automatically verifying self-consistent mpi performance guidelines. arXiv:1606.00215, https://doi.org/10.48550/arXiv.1606.00215

  8. Huang X, Ramos FA, Deng Y (2022) Optimal circulant graphs as low-latency network topologies. J Supercomput 78(11):13491–13510. https://doi.org/10.1007/s11227-022-04396-5

    Article  Google Scholar 

  9. Sun X-H et al (2003) Improving the performance of mpi derived datatypes by optimizing memory-access cost. In: 2003 Proceedings IEEE International Conference on Cluster Computing. IEEE, pp 412–419. https://doi.org/10.1109/CLUSTR.2003.1253341

  10. Andoh Y, Ichikawa S-I, Sakashita T, Yoshii N, Okazaki S (2021) Algorithm to minimize mpi communications in the parallelized fast multipole method combined with molecular dynamics calculations. J Comput Chem 42(15):1073–1087. https://doi.org/10.1002/jcc.26524

    Article  CAS  PubMed  Google Scholar 

  11. Suresh KK, Ramesh B, Ghazimirsaeed SM, Bayatpour M, Hashmi J, Panda DK (2020) Performance characterization of network mechanisms for non-contiguous data transfers in mpi. In: 2020 IEEE International parallel and distributed processing symposium workshops (IPDPSW). IEEE, pp 896–905. https://doi.org/10.1109/IPDPSW50202.2020.00150

  12. Awan AA, Manian KV, Chu C-H, Subramoni H, Panda DK (2019) Optimized large-message broadcast for deep learning workloads: Mpi, mpi+nccl, or nccl2? Parallel Comput 85:141–152. https://doi.org/10.1016/j.parco.2019.03.005

    Article  Google Scholar 

  13. White S, Kale LV (2020) Optimizing point-to-point communication between adaptive mpi endpoints in shared memory. Concurr Comput Pract Exp 32(3):4467–4479. https://doi.org/10.1002/cpe.4467

    Article  Google Scholar 

  14. Feng G, Dong D, Lu Y (2022) Optimized mpi collective algorithms for dragonfly topology. In: Proceedings of the 36th ACM International Conference on Supercomputing. Association for Computing Machinery, New York, pp 1–11. https://doi.org/10.1145/3524059.3532380

  15. Kang Q, Lee S, Hou K, Ross R, Agrawal A, Choudhary A, Liao W-K (2020) Improving mpi collective i/o for high volume non-contiguous requests with intra-node aggregation. IEEE Trans Parallel Distrib Syst 31(11):2682–2695. https://doi.org/10.1109/TPDS.2020.3000458

    Article  Google Scholar 

  16. Wagle B, Kellar S, Serio A, Kaiser H (2018) Methodology for adaptive active message coalescing in task based runtime systems. In: 2018 IEEE international parallel and distributed processing symposium workshops (IPDPSW). IEEE, pp 1133–1140. https://doi.org/10.1109/IPDPSW.2018.00173

  17. The Ohio State University N-BCLN (2020) MVAPICH: MPI over InfiniBand, Omni-Path, Ethernet/iWARP, RoCE, and Slingshot. http://mvapich.cse.ohio-state.edu/benchmarks/

  18. Gallardo E, Vienne J, Fialho L, Teller P, Browne J (2015) Mpi advisor: a minimal overhead tool for mpi library performance tuning. In: Proceedings of the 22Nd European MPI users’ group meeting, pp 1–10. https://doi.org/10.1145/2802658.2802667

  19. Du Q, Huang H (2022) Mpi parameter optimization during debugging phase of hpc system. J Supercomput 78:1696–1711. https://doi.org/10.1007/s11227-021-03939-6

    Article  Google Scholar 

  20. Forejt V, Kroening D, Narayanaswamy G, Sharma S (2014) Precise predictive analysis for discovering communication deadlocks in mpi programs. In: FM 2014: Formal Methods: 19th International symposium, Singapore, May 12–16, 2014. Proceedings 19. Springer, Cham, pp 263–278. https://doi.org/10.1007/978-3-319-06410-9_19

  21. Forum M (2012) MPI: A Message-Passing Interface Standard Version 3.0. https://www.mpi-forum.org/docs/mpi-3.0/mpi30-report.pdf

  22. Ghazimirsaeed SM, Mirsadeghi SH, Afsahi A (2020) Communication-aware message matching in mpi. Concurr Comput Pract Exp 32(3):4862–4879. https://doi.org/10.1002/cpe.4862

    Article  Google Scholar 

Download references

Acknowledgements

This work was supported by the National Key Research and Development Program of China, Grant/Award Number: 2016YFB0201100. And R &D and application of key technologies of independent and controllable computing power network, Grant/Award Number: 2022JBZ01-01.

Funding

This work was supported by the National Key Research and Development Program of China, Grant/Award Number: 2016YFB0201100. And R &D and application of key technologies of independent and controllable computing power network, Grant/Award Number: 2022JBZ01-01.

Author information

Author notes

  1. Jubin Wang, Yuan Zhuang and Yunhui Zeng have contributed equally to this work.

Authors and Affiliations

  1. Key Laboratory of Computing Power Network and Information Security, Ministry of Education, Shandong Computer Science Center (National Supercomputer Center in Jinan), Qilu University of Technology (Shandong Academy of Sciences), Jinan, China

    Jubin Wang, Yuan Zhuang & Yunhui Zeng

  2. Shandong Provincial Key Laboratory of Computer Networks, Shandong Fundamental Research Center for Computer Science, Jinan, China

    Jubin Wang, Yuan Zhuang & Yunhui Zeng

Authors
  1. Jubin Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Yuan Zhuang
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Yunhui Zeng
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

Author 1 was involved in paper writing, data testing and analysis, scanning module development and design, and data testing. Author 2 helped in software package design and development, as well as POP testing, results and analysis. Author 3 (corresponding author) contributed to overall paper conception, participation in the entire development and testing process.

Corresponding author

Correspondence to Yunhui Zeng.

Ethics declarations

Conflict of interest

All authors disclosed no relevant relationships.

Ethics approval

This article does not contain any studies involving humans or animals.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Wang, J., Zhuang, Y. & Zeng, Y. A transmission optimization method for MPI communications. J Supercomput 80, 6240–6263 (2024). https://doi.org/10.1007/s11227-023-05699-x

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11227-023-05699-x

Keywords

Search

Navigation