Towards dynamic adaptability support for the master-worker paradigm in component based applications

Françoise André (1) , Hinde Lilia Bouziane (1) , Jérémy Buisson (1) , Jean-Louis Pazat (1) , Christian Pérez (1)

1 PARIS - Programming distributed parallel systems for large scale numerical simulation

When executing scientific applications, resources that may be used can vary from multi-core processors to grids. Therefore, abstracting the programming model enables portability on various resource infrastructures. Furthermore, software component technology appears to be a very promising approach to deal with the growing complexity of scientific applications. Hence, we proposed a model to improve the support of master-worker paradigm in component models. Capitalizing on our experience of adaptability frameworks, we propose to enhance our model so that master-worker applications can adapt at runtime to varying conditions. This paper studies how to transparently introduce adaptability in our model for master-worker applications, what impact it has on the model, and what requirements it expects from the adaptability framework.

Mots clés: software components; Grid; master-worker; dynamic evolution; adaptability framework

Domaines: Calcul parallèle, distribué et partagé [cs.DC]

Françoise André, Hinde Lilia Bouziane, Jérémy Buisson, Jean-Louis Pazat, Christian Pérez. Towards dynamic adaptability support for the master-worker paradigm in component based applications. CoreGRID Symposium, Aug 2007, Rennes, France. pp.117, ⟨10.1007/978-0-387-72498-0_11⟩. ⟨hal-00498872⟩