Distributed management of distributed systems
May 15, 15.00 (Aula C)
May 20, 15.00 (Aula F)
Palazzo delle Scienze - Cagliari
INRIA - IRISA Rennes
Abstract. In these lectures, we examine several formalisms that have been introduced to model dynamic systems with concurrent behaviors. Concurrency represents the ability for a system to perform several actions at a time, and it is a central characteristic of most large scale systems that are of common use today, either at the physical level (embedded systems, clouds or networks of machines, telecommunication networks) or at the software level (distributed applications).
Verifying and managing these systems requires formalisms that capture the concurrency of events, and standard automata are inappropriate for that. This lecture will examine first attempts through Mazurkievicz traces, then we will analyze networks of automata with an appropriate semantics to capture concurrency, and we will finish with Petri nets.
The stress will be on distributed/modular approaches to the verification/management of such systems. Algebraic properties of the trajectory space of these systems will be detailed, that enable one to analyse these systems by parts. Problems like distributed diagnosis or distributed planning will be examined, as illustrations of these distributed management approaches.