Une heuristique d'ordonnancement et de distribution tolérante aux pannes pour systèmes temps-réel embarqués

Abstract

In this paper, we present a new task scheduling heuristic on distributed multiprocessor architectures, which allows to generate a static distributed fault tolerant schedule, minimising the length of the whole generated schedule. The heuristic that we propose is based on the active replication of the tasks, where each task is replicated at least on Npf+1 different processors to tolerate Npf processor failures. Through a detailed example, we show the techniques used to minimise the length of the schedule and to tolerate the failures of processors. A study by simulations shows the efficiency of our method compared to other heuristics found in the literature.

BibTeX entry

@InProceedings{GKS03,
  author = 	 {A. Girault and C. Lavarenne and Y. Sorel},
  title = 	 {Une heuristique d'ordonnancement et de distribution
                  tolérante aux pannes pour systèmes temps-réel embarqués},
  booktitle = 	 {Mod\'{e}lisation des Syst\`{e}mes R\'{e}actifs, MSR'03},
  year =	 {2003},
  address =	 {Metz, France},
  month =	 oct,
  publisher =	 {Hermes},
}