Capacity Planning of Supercomputers: Simulating MPI Applications at Scale

Tom Cornebize 1
1 POLARIS - Performance analysis and optimization of LARge Infrastructures and Systems
Inria Grenoble - Rhône-Alpes, LIG - Laboratoire d'Informatique de Grenoble
Abstract : The capacity planning of supercomputers is a complex but important task. The best combination of hardware for the given budget has to be selected. Such choices are often guided by some common rules, elaborated after years of trial and error. The rapid development of the field, however, makes this approach costly and ineffective. We believe that experts could benefit from accurate and efficient simulators to design the next generations of supercomputers. In this work, we simulate a complex MPI application at the scale of the current top500 supercomputers, using the Simgrid simulation toolkit. We present the optimization developed to enable this kind of large scale executions, we demonstrate the soundness of the resulting simulation and we illustrate the new possibilities of study.
Liste complète des métadonnées

Cited literature [18 references]  Display  Hide  Download

https://hal.inria.fr/hal-01544827
Contributor : Tom Cornebize <>
Submitted on : Thursday, June 22, 2017 - 10:58:29 AM
Last modification on : Thursday, October 11, 2018 - 8:48:05 AM

Identifiers

  • HAL Id : hal-01544827, version 1

Citation

Tom Cornebize. Capacity Planning of Supercomputers: Simulating MPI Applications at Scale. Distributed, Parallel, and Cluster Computing [cs.DC]. 2017. ⟨hal-01544827⟩

Share

Metrics

Record views

509

Files downloads

252