The large use, in ourdays, of networks of small unreliable devices with unreliable connectivity needs researchs for the design of reliable, scalable and safe distributed algorithms.

**Reliability of distributed algorithms**: using local computations paradigm (a model defined in LaBRI), we aim to prove formally (using Coq for example) that the formal specification of the algorithms is valid.**Fault-tolerant algorithmes**: distributed systems are increasingly subject to disturbances due to permanent and/or transient failures of nodes and to their mobility. We are interested on the design of fault-tolerant algorithms. This include paradigms like wait-freedom, obstruction-freedom, self-stabilization, failure detector ... .**Models for huge and highly dynamic distributed networks**: we are wistening, in few last years, the emergence of topological properties that describe the aspects of physical and virtual networks (small-world phenomena, ...). Unfortunately, there is a lack of efficient models which can capture these properties and which can be easily used to design polynomial distributed algorithms to solve classical problems.**Distributed big data**: this theme aims to design virtual peer to peer networks and routing algorithms. It also develop efficient requests execution (exploration, cartography, rendez-vous, ...) using mobile agents, broadcasting and distributed computing in large dynamic networks. We design high speed solutions using small amount of memory, and try to ensure, as far as possible, a performances comparable to that of centralized solutions.**Local computation**: this axis aims to provide tools to encode, study and teach distributed algorithms. We are mainly interested on classical problems like election, numbering, spanning tree computation, termination detection, consensus ... The power of the tools are studied depending on the computation rules which are used and on the initial knowledge of the nodes in the system.**Randomized distributed algorithms**: under some particular hypotheses (anonymous networks, ...), there is no deterministic solution to solve some classical problems from distributed algorithms theory (election, colouring, ...). Randomized algorithms can be used to solve efficiently such problems. These algorithms may be simple to design, however, their analysis is more challenging and needs the use of sophisticated methods from combinatorics, algorithms analysis, ... We are particularly interested on the design of optimal solutions from bit complexity point of view.

Récupéré sur http://algodist.labri.fr/index.php/Main/HomePage

Page mise à jour le 17 octobre 2017 à 20h33