Skip to Main content Skip to Navigation

Modularité et symétrie pour les systèmes répartis; application au langage CSP

Abstract : Numerical criteria are usually considered to evaluate distributed systems. They are mainly based on information transfers within computations. We show those criteria are not satisfactory in evaluating the degree of distributedness of distributed algorithms. Specifically designed non-numerical criteria are needed.

Genericity expresses that processors have initially no knowledge about the network they are embedded in. Symmetry expresses that processors with topologically equivalent positions in the network, play equivalent rôles within computations, too.

We define those properties in the framework of Hoare's language CSP of Communicating Sequential Processes. We give a syntactic definition for genericity. We show that such a definition is not appropriate for symmetry. A semantic definition is needed.

We consider then broadcasting and leader-finding algorithms for networks of communicating processes in CSP. We study the existence of generic and symmetric solutions. Many results, positive as well as negative, are carried out. This leads to a precise evaluation of the expressive power of CSP. In particular, we show that output guards cannot be implemented by means of input guards only, as long as symmetry is preserved.

These results are finally applied to describe a generic, symmetric and bounded solution to Francez' problem of distributed termination detection.
Complete list of metadata
Contributor : Luc Bougé Connect in order to contact the contributor
Submitted on : Saturday, September 12, 2009 - 10:28:19 PM
Last modification on : Thursday, March 17, 2022 - 10:08:24 AM
Long-term archiving on: : Tuesday, June 15, 2010 - 11:35:56 PM


  • HAL Id : tel-00416184, version 1



Luc Bougé. Modularité et symétrie pour les systèmes répartis; application au langage CSP. Génie logiciel [cs.SE]. Université Paris-Diderot - Paris VII, 1987. Français. ⟨tel-00416184⟩



Record views


Files downloads