Ph.D
Group : Parallelism
Observing self-stabilization
Starts on 09/01/2002
Advisor : ROZOY, Brigitte
[Joffroy BEAUQUIER]
Funding : MNRT
Affiliation : Université Paris-Saclay
Laboratory : LRI
Defended on 15/12/2005, committee :
Joffroy Beauquier
Véronique Benzaken
Béatrice Bérard
Yves Métivier
Brigitte Rozoy
André Schiper
Research activities :
- Verification
- Distributed algorithms
- Self-stabilisation
- Design
- Fault Tolerance
Abstract :
Distributed systems have two main characteristics: they are complex
and subject to failures. So, the verification and the study of fault
tolerance in such systems are two major issues. In this thesis, we
propose a model-checking verification technique for
distributed systems. We present and prove an algorithm, based on
partial orders, which builds a small subset of the whole set of states
of a distributed system. In the reduced generated graph, it is still
possible to verify stable properties of the considered system. We
then study self-stabilizing systems which are fault tolerant
distributed systems. Self-stabilizing systems are systems which, from
any initialization, eventually behaves correctly, regarding to their
specifications. The drawback of such systems is that they cannot
determine whether or not they verify their specifications. In this
thesis we propose a new model, in which the system can decide if it
verifies its specification by introducing a new abstraction called
observer. With this model, we prove that if there exists a synchronous
self-stabilizing distributed solution for some problem in a
distinguished network, then there exists a synchronous
self-stabilizing distributed solution for the same problem in the same
network which accepts an observer. Finally, we introduce the notion of
a probabilistic observer and we prove that such an observer allows to
decide for a larger class of self-stabilizing systems than
deterministic observers.
More information: http://info.iut-bm.univ-fcomte.fr/staff/pilard/publication/publication.php