Description
Date depot: 13 janvier 2020
Titre: Statistical-oriented predictability of cyber-physical systems
Directeur de thèse:
Avner BAR-HEN (CEDRIC)
Directrice de thèse:
Liliana CUCU-GROSJEAN (Inria-Paris (ED-130))
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Systèmes et architectures intégrés matériel-logiciel
Resumé:
What a defibrillator, a car or an airplane have in common? They all share the same property of being a cyber-physical system (CPS). A CPS is an integration of computation with physical processes whose behavior is defined by both cyber and physical parts of the system [1]. An increased pressure is pushing the designers of these systems to make them as autonomous as possible. This pressure is translated at program level by the implementation of smarter algorithms that are tested on the only processors available on the market today: the multicore processors. Indeed, due to the pioneer utilization of multicore processors by the smartphone market as well as the impressive expansion of this later market, the microprocessor industry has evolved towards general purpose processors with complex architectures that are not time predictable. Their lack of time predictability is due to features like several cores, multiple levels of caches and pipelines, speculative branching, communicating through shared memory or/and through a network on chip, etc. While smartphone clients are willing to charge often their phones or to re-boot regularly their applications to compensate for the bad design of the phones on multicore processors, the rest of the CPS industry is facing the open problem of time predictability of programs on multicore processors if they want to provide to their clients applications that are stable and have low-energy consumption.
Unfortunately bounding the execution time of a program on multicore processors is known to be an open problem. For instance, there is no program ensuring the autonomic behavior of an airplane executed on multicore processors today because of the unbounded execution time of the programs on such processors [2]. This research project deals with the proposition of appropriate statistical models bounding the execution times probability distributions of programs on multicore processors.
[1] R. Baheti and H. Gill. Cyber-physical systems. IEEE, 2011
[2] P. Parkinson. Towards avionics safety certification on multi-core processor architectures. TechTime, Elec-
tronics and Technology News, 2016.
Doctorant.e: Zagalo Kevin