Article

On the Convergence of Malleability and the HPC PowerStack: Exploiting Dynamism in Over-Provisioned and Power-Constrained HPC Systems

Authors:

A. Isaías Comprés,

Martin SchulzAuthors Info & Claims

High Performance Computing. ISC High Performance 2022 International Workshops: Hamburg, Germany, May 29 – June 2, 2022, Revised Selected Papers

Pages 206 - 217

https://doi.org/10.1007/978-3-031-23220-6_14

Published: 04 January 2023 Publication History

Abstract

Recent High-Performance Computing (HPC) systems are facing important challenges, such as massive power consumption, while at the same time significantly under-utilized system resources. Given the power consumption trends, future systems will be deployed in an over-provisioned manner where more resources are installed than they can afford to power simultaneously. In such a scenario, maximizing resource utilization and energy efficiency, while keeping a given power constraint, is pivotal. Driven by this observation, in this position paper we first highlight the recent trends of resource management techniques, with a particular focus on malleability support (i.e., dynamically scaling resource allocations/requirements for a job), co-scheduling (i.e., co-locating multiple jobs within a node), and power management. Second, we consider putting them together, assess their relationships/synergies, and discuss the functionality requirements in each software component for future over-provisioned and power-constrained HPC systems. Third, we briefly introduce our ongoing efforts on the integration of software tools, which will ultimately lead to the convergence of malleability and power management, as it is designed in the HPC PowerStack initiative.

References

[1]

Deep-sea: Programming environment for european exascale systems. https://www.deep-projects.eu/, Accessed 25 Apr 2022

[2]

The hpc powerstack. https://hpcpowerstack.github.io/index.html, LNCS Accessed 25 Apr 2022

[3]

Regale: Open architecture for exascale supercomputers. https://regale-project.eu/, Accessed 25 Apr 2022

[4]

Top 500. https://www.top500.org/statistics/list/, Accessed 28 Feb 2022

[5]

Ahn DH et al. Flux: overcoming scheduling challenges for exascale workflows Future Gener. Comput. Syst. 2020 110 202-213

[6]

Aupy, G., et al.: Co-scheduling HPC workloads on cache-partitioned CMP platforms. In: CLUSTER, pp. 348–358 (2018)

[7]

Bartolini, A., et al.: A pulp-based parallel power controller for future exascale systems. In: ICECS, pp. 771–774 (2019)

[8]

Bhadauria, M., et al.: An approach to resource-aware co-scheduling for CMPs. In: ICS, pp. 189–199 (2010)

[9]

Borghesi, A., et al.: Examon-x: a predictive maintenance framework for automatic monitoring in industrial iot systems. IEEE Internet Things J. (2021)

[10]

Breitbart, J., et al.: Case study on co-scheduling for HPC applications. In: ICPPW, pp. 277–285 (2015)

[11]

Breitbart, J., et al.: Dynamic co-scheduling driven by main memory bandwidth utilization. In: CLUSTER, pp. 400–409 (2017)

[12]

Breslow, A.D., et al.: Enabling fair pricing on hpc systems with node sharing. In: SC (2013)

[13]

Capit, N., et al.: A batch scheduler with high level components. In: CCGrid, vol. 2, pp. 776–783 (2005)

[14]

Castain RH et al. Pmix: process management for exascale environments Parallel Comput. 2018 79 9-29

[15]

Cesarini D et al. Countdown slack: a run-time library to reduce energy footprint in large-scale mpi applications IEEE TPDS 2020 31 11 2696-2709

[16]

Cochran, R., et al.: Pack & cap: adaptive dvfs and thread packing under power caps. In: MICRO, pp. 175–185 (2011)

[17]

Comprés, I., et al.: Infrastructure and api extensions for elastic execution of mpi applications, pp. 82–97. EuroMPI (2016)

[18]

Corbalan, J., et al.: EAR: energy management framework for supercomputers. In: Barcelona Supercomputing Center (BSC) Working paper (2019)

[19]

D’Amico, M., et al.: Holistic slowdown driven scheduling and resource management for malleable jobs. In: ICPP (2019)

[20]

Esmaeilzadeh, H., et al.: Dark silicon and the end of multicore scaling. In: ISCA, pp. 365–376 (2011)

[21]

Feitelson, D.G., et al.: Toward convergence in job schedulers for parallel supercomputers. In: JSSPP, pp. 1–26 (1996)

[22]

Hennessy, J., Patterson, D.: A new golden age for computer architecture: domain-specific hardware/software co-design, enhanced. In: ISCA (2018)

[23]

Kale, L.V., et al.: A malleable-job system for timeshared parallel machines. In: CCGRID, pp. 230–230 (2002)

[24]

Mo-Hellenbrand, A., et al.: A large-scale malleable tsunami simulation realized on an elastic mpi infrastructure. In: CF, pp. 271–274 (2017)

[25]

Netti, A., et al.: From facility to application sensor data: modular, continuous and holistic monitoring with dcdb. In: SC, pp. 1–27 (2019)

[26]

Patki, T., et al.: Exploring hardware overprovisioning in power-constrained, high performance computing. In: ICS, pp. 173–182 (2013)

[27]

Patki, T., et al.: Practical resource management in power-constrained, high performance computing. In: HPDC, pp. 121–132 (2015)

[28]

Sakamoto, R., et al.: Analyzing resource trade-offs in hardware overprovisioned supercomputers. In: IPDPS, pp. 526–535 (2018)

[29]

Sarood, O., et al.: Maximizing throughput of overprovisioned HPC data centers under a strict power budget. In: SC, pp. 807–818 (2014)

[30]

Schreiber, M., et al.: Invasive compute balancing for applications with hybrid parallelization. In: SBAC-PAD, pp. 136–143 (2013)

[31]

Scogland, T.R., et al.: A power-measurement methodology for large-scale, high-performance computing. In: ICPE, pp. 149–159 (2014)

[32]

Shalf J The future of computing beyond moore’s law Phil. Trans. Roy. Soc. A 2020 378 2166 20190061

[33]

Utrera, G., et al.: A job scheduling approach for multi-core clusters based on virtual malleability. In: Euro-Par, pp. 191–203 (2012)

[34]

Vigouroux, X., et al.: Towards energy consumption application profiling with bull energy software. https://prace-ri.eu/wp-content/uploads/PRACE-at-SC17-Ludovic-Sauge.pdf, Accessed 14 Mar 2022

[35]

Yoo, A.B., et al.: Slurm: simple linux utility for resource management. In: JSSPP, pp. 44–60 (2003)

[36]

Zhu, Q., et al.: Co-run scheduling with power cap on integrated CPU-GPU systems. In: IPDPS, pp. 967–977 (2017)

Cited By

Chadha MArima ERaoofy AGerndt MSchulz M(2023)Sustainability in HPC: Vision and OpportunitiesProceedings of the SC '23 Workshops of The International Conference on High Performance Computing, Network, Storage, and Analysis10.1145/3624062.3624271(1876-1880)Online publication date: 12-Nov-2023
https://dl.acm.org/doi/10.1145/3624062.3624271
Wilson DAcun FJana SArdanaz FEastep JPaschalidis ICoskun A(2023)An End-to-End HPC Framework for Dynamic Power ObjectivesProceedings of the SC '23 Workshops of The International Conference on High Performance Computing, Network, Storage, and Analysis10.1145/3624062.3624262(1801-1811)Online publication date: 12-Nov-2023
https://dl.acm.org/doi/10.1145/3624062.3624262
Besnard JTarraf ABarthélemy CCascajo AJeannot EShende SWolf F(2023)Towards Smarter Schedulers: Molding Jobs into the Right Shape via Monitoring and ModelingHigh Performance Computing10.1007/978-3-031-40843-4_6(68-81)Online publication date: 21-May-2023
https://dl.acm.org/doi/10.1007/978-3-031-40843-4_6
Show More Cited By

Index Terms

On the Convergence of Malleability and the HPC PowerStack: Exploiting Dynamism in Over-Provisioned and Power-Constrained HPC Systems

Index terms have been assigned to the content through auto-classification.

Recommendations

Demand-aware power management for power-constrained HPC systems
CCGRID '16: Proceedings of the 16th IEEE/ACM International Symposium on Cluster, Cloud, and Grid Computing

As limited power budget is becoming one of the most crucial challenges in developing supercomputer systems, hardware overprovisioning which installs larger number of nodes beyond the limitations of the power constraint determined by Thermal Design Power ...
Power Tuning HPC Jobs on Power-Constrained Systems
PACT '16: Proceedings of the 2016 International Conference on Parallel Architectures and Compilation

As we approach the exascale era, power has become a primary bottleneck. The US Department of Energy has set a power constraint of 20MW on each exascale machine. To be able achieve one exaflop under this constraint, it is necessary that we use power ...
Parallel job scheduling for power constrained HPC systems

Power has become the primary constraint in high performance computing. Traditionally, parallel job scheduling policies have been designed to improve certain job performance metrics when scheduling parallel workloads on a system with a given number of ...

Comments

Information & Contributors

Information

Published In

cover image Guide Proceedings

High Performance Computing. ISC High Performance 2022 International Workshops: Hamburg, Germany, May 29 – June 2, 2022, Revised Selected Papers

May 2022

398 pages

ISBN:978-3-031-23219-0

DOI:10.1007/978-3-031-23220-6

Editors:
Hartwig Anzt
University of Tennessee, Knoxville, TN, USA
,
Amanda Bienz
University of New Mexico, Albuquerque, NM, USA
,
Piotr Luszczek
University of Tennessee, Knoxville, TN, USA
,
Marc Baboulin
Université Paris-Saclay, Orsay, France

© Springer Nature Switzerland AG 2022.

Publisher

Springer-Verlag

Berlin, Heidelberg

Publication History

Published: 04 January 2023

Author Tags

Qualifiers

Article

Contributors

Other Metrics

View Article Metrics

Bibliometrics & Citations

Bibliometrics

Article Metrics

4
Total Citations
View Citations
0
Total Downloads

Downloads (Last 12 months)0
Downloads (Last 6 weeks)0

Reflects downloads up to 24 Jan 2025

Other Metrics

View Author Metrics

Citations

Cited By

Chadha MArima ERaoofy AGerndt MSchulz M(2023)Sustainability in HPC: Vision and OpportunitiesProceedings of the SC '23 Workshops of The International Conference on High Performance Computing, Network, Storage, and Analysis10.1145/3624062.3624271(1876-1880)Online publication date: 12-Nov-2023
https://dl.acm.org/doi/10.1145/3624062.3624271
Wilson DAcun FJana SArdanaz FEastep JPaschalidis ICoskun A(2023)An End-to-End HPC Framework for Dynamic Power ObjectivesProceedings of the SC '23 Workshops of The International Conference on High Performance Computing, Network, Storage, and Analysis10.1145/3624062.3624262(1801-1811)Online publication date: 12-Nov-2023
https://dl.acm.org/doi/10.1145/3624062.3624262
Besnard JTarraf ABarthélemy CCascajo AJeannot EShende SWolf F(2023)Towards Smarter Schedulers: Molding Jobs into the Right Shape via Monitoring and ModelingHigh Performance Computing10.1007/978-3-031-40843-4_6(68-81)Online publication date: 21-May-2023
https://dl.acm.org/doi/10.1007/978-3-031-40843-4_6
Taboada HPereira RJaeger JBesnard J(2023)Towards Achieving Transparent Malleability Thanks to MPI Process VirtualizationHigh Performance Computing10.1007/978-3-031-40843-4_3(28-41)Online publication date: 21-May-2023
https://dl.acm.org/doi/10.1007/978-3-031-40843-4_3

View Options

View options

Figures

Tables

Media

View Table of Conten