A Case Study on PMIx-Usage for Dynamic Resource Management - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

A Case Study on PMIx-Usage for Dynamic Resource Management

Résumé

With the increasing scale of HPC supercomputers efficient resource utilization on such systems becomes even more important. In this context, dynamic resource management is a very active research field, as it is expected to improve several metrics of resource utilization on HPC systems, such as job throughput and energy efficiency. However, dynamic resource management is complex and requires significant changes to various layers of the software stack including resource- and process management, programming models and applications. So far, approaches for resource management are often specific to a particular implementation of the resource management and process management software, thus hindering interoperability, composability and comparability of such approaches. In this paper, we discuss the usage of the Process Management Interface - Exascale (PMIx) Standard for interactions between the process manager and the resource manager. We describe an architecture that allows the resource manager to connect to the process manager as PMIx Tool to have access to a set of PMIx services useful for resource management. In a concrete case-study we connect a python- and PMIx-based resource manager to PRRTE and assess the applicability of this architecture for debugging and exploration of dynamic resource management techniques. We conclude that a PMIx-based architecture can simplify the process of exploring new dynamic and disruptive resource management mechanisms while improving composability.

Domaines

Informatique
Fichier principal
Vignette du fichier
paper_hpcmall_FINAL_PREPRINT.pdf (318.31 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04214980 , version 1 (06-02-2024)

Licence

Paternité

Identifiants

Citer

Dominik Huber, Martin Schreiber, Martin Schulz. A Case Study on PMIx-Usage for Dynamic Resource Management. High Performance 2023 - 38th International Conference, ISC High Performance, May 2023, Hambourg, Germany. pp.42-55, ⟨10.1007/978-3-031-40843-4_4⟩. ⟨hal-04214980⟩
61 Consultations
19 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More