Section: Partnerships and Cooperations

International Initiatives

DISTOL

• Title: Distributed systems, stochastic models and logics

• International Partner (Institution - Laboratory - Researcher):

  • CMI (India) - Madhavan Mukund

• Start year: 2013

• See also: http://www.irisa.fr/sumo/DISTOL/

• The context of this project is formal modeling, and analysis of behaviors of distributed systems. We want to address verification and supervision of distributed systems through formal modeling and automated reasoning on models. By distributed systems, we mean software architectures made of several independent communicating entities. In the 90?s the kind of system addressed was mainly telecommunication protocols. Nowadays, distributed systems are frequently web-based systems such as Web Services, but several aspects of distributed systems can be found in biological applications. Within this context, a challenge is to propose formal tools with potential applications to real systems. We want to address this challenge along three main axes: The first one is realism of models. Models are often an abstraction of real systems. We want to build and study properties of models that are close enough from their implementations, and with robust properties. By robustness, we mean that properties checked on a model (for instance safety properties) should still hold for implementations of this model. The second one is quantitative analysis of systems. Rather that considering boolean answers to formal properties, one can consider the probability that such property holds on a run of the system, and return answers of probabilistic form (“almost surely, a call to a service is successful”) or quantitative (“the average failure rate is lower than 0.01”). One possibility to obtain a probability is to compute its exact value. Such questions have answers for markovian models and some quantitative logics (PCTL). However, such computations are expensive, and one can divide the problem into sub-components at the cost of some approximation. We plan to develop efficient algorithms for quantitative analysis of systems. The third one is unification of control theories. There are many proposals for supervisory control, including distributed control with communications. However, none of them seems fully satisfactory. We want to consider connections between control theory, epistemic reasoning (which seems to solve some problems raised by communications between local supervisors), and game theory (which emphasizes the notion of goal to be achieved in a problem), and give a unified framework for supervision of distributed systems.

The team collaborates on runtime enforcement with the group of Prof. Stavros Tripakis (http://users.ics.aalto.fi/stavros/ ) at Aalto University (Finland), where our former PhD student Srinivas Pinisetty is doing a Post-doc.

In the context of LIRIMA, the Inria International Lab in Africa, we have strong collaborations with University of Yaoundé I on an artifact-centric model of workflow system based on guarded attribute grammars. In particular with the co-supervision of the PhD thesis of Robert Nsaibirni.

We collaborate with Laurie Ricker (Mount Allison University, Canada) on the control of distributed systems and the enforcement of opacity

AVeRTS is an Indo-French project on the algorithmic verification of real-time systems. The project is funded by CNRS on the french side, and by DST on the Indian side, under the CEFIPRA - Indo-French Program in ICST 2014-2016. From SUMO, Nathalie Bertrand and Blaise Genest are involved and contribute on stochastic games. In the context of this project, Miheer Dewaskar, a CMI (Chennai Mathematical Institute) master student did an internship in our team on the control of a population of Markov decision processes.