The Omniscient Garbage Collector : a Resource Analysis Framework - Archive ouverte HAL
Communication Dans Un Congrès Année : 2014

The Omniscient Garbage Collector : a Resource Analysis Framework

Aurélien Deharbe
  • Fonction : Auteur
  • PersonId : 954864

Résumé

The notion of resource plays a central role in concurrent systems. In its purest form a resource is simply a unique identity one can create, use and ultimately destruct. In this paper we propose a simple yet effective characterization of resource usages and develop for it a complete analysis framework. We address qualitative issues such as the classification of resources and whether two systems exhibit similar patterns of resource usages -- namely equivalent resource profiles. From the quantitative point of view, we develop the omniscient garbage collector (OGC), which decides precisely when a resource can be reclaimed or reused. This allows to bound precisely the number of resources consumed by a given system. To illustrate the approach, we study experimentally the resource consumption of pi-calculus processes using a prototype analysis tool. We propose two different resource abstractions for pi-processes: one based on the labelled transitions for open systems, and another one for closed systems. The latter notably provides a refined view of behaviors, less opaque than reductions. Beyond this experiment, the proposed framework is quite generic and can apply to many different formalisms and situations.
Fichier principal
Vignette du fichier
ogc_acsd14_submission.pdf (221.16 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-00973747 , version 1 (04-04-2014)

Identifiants

Citer

Aurélien Deharbe, Frédéric Peschanski. The Omniscient Garbage Collector : a Resource Analysis Framework. ACSD 2014 - 14th International Conference on Application of Concurrency to System Design, Jun 2014, La Marsa, Tunisia. pp.102-111, ⟨10.1109/ACSD.2014.18⟩. ⟨hal-00973747⟩
182 Consultations
199 Téléchargements

Altmetric

Partager

More