An RMS Architecture for Efficiently Supporting Complex-Moldable Applications

Cristian Klein 1 Christian Pérez 1
1 GRAAL - Algorithms and Scheduling for Distributed Heterogeneous Platforms
Inria Grenoble - Rhône-Alpes, LIP - Laboratoire de l'Informatique du Parallélisme
Abstract : High-performance scientific applications are becoming increasingly complex, in particular because of the coupling of parallel codes. This results in applications having a complex structure, characterized by multiple deploy-time parameters, such as the number of processes of each code. In order to optimize the performance of these applications, the parameters have to be carefully chosen, a process which is highly resource dependent. However, the abstractions provided by current Resource Management Systems (RMS) - either submitting rigid jobs or enumerating a list of moldable configurations - are insufficient to efficiently select resources for such applications. This paper introduces CooRM, an RMS architecture that delegates resource selection to applications while still keeping control over the resources. The proposed architecture is evaluated using a simulator which is then validated with a proof-of-concept implementation on Grid'5000. Results show that such a system is feasible and performs well with respect to scalability and fairness.
Complete list of metadatas

Cited literature [21 references]  Display  Hide  Download

https://hal.inria.fr/inria-00606771
Contributor : Cristian Klein <>
Submitted on : Thursday, July 7, 2011 - 10:42:14 AM
Last modification on : Friday, April 20, 2018 - 3:44:23 PM
Long-term archiving on : Saturday, October 8, 2011 - 2:21:53 AM

File

main.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : inria-00606771, version 1

Collections

Citation

Cristian Klein, Christian Pérez. An RMS Architecture for Efficiently Supporting Complex-Moldable Applications. IEEE International Conference on High Performance Computing and Communications, Sep 2011, Banff, Alberta, Canada. ⟨inria-00606771⟩

Share

Metrics

Record views

315

Files downloads

345