Coopération

Séminaires


< retour aux séminaires

Yvon Savaria (*, presenter) Ghaith Bany Hamad (*), Anuarul Hoque (+), and Otmane Ait-Mohamed (+), This seminar presents some fruits of a collaboration between (+) Concordia University and (*) Polytechnique Montréal

Theme: Towards Multilevel Formal Probabilistic Modeling, Analysis, and Estimation of Soft Error and High-Level SEU Sensitivity Analysis Using Probabilistic Model Checking
Date: Le jeudi 11 juin 2015 à 10h, Laboratoire TIMA - Salle T312

Biography

Le professeur Savaria, a une longue expérience de la conception des systèmes électroniques exploitant les techniques d’intégration et les méthodes de conception modernes qui s’y appliquent. Actif dans le domaine depuis plus de 30 ans, titulaire d’une chaire de recherche du Canada à Polytechnique Montréal, il a publié avec ses collaborateurs plus de 500 articles scientifiques, 15 brevets et un livre. Plus de 150 étudiants chercheurs ont rédigé des thèses (PhD) et mémoires (MScA) sous sa direction et au cours de sa carrière, plus de 100 circuits intégrés ont été conçus et fabriqués en exploitant une dizaine de technologies. Son approche système l’amène à considérer les méthodes de conception et les moyens tant numériques qu’analogiques et mixtes comme des moyens de satisfaire des requis système. Il a notamment plusieurs contributions récentes axées vers la conception d’accélérateurs de calcul et d’ASIP. Il s’est intéressé à diverses classes d’applications en télécommunications, matériel d’infrastructure de réseau, traitement vidéo, prototypage rapide et aérospatial. Il s'intéresse notamment à la conception de systèmes électroniques fiable pour des applications critiques.

Il est membre fondateur de RESMIQ, un regroupement stratégique Québéçois qui rassemble 50 professeurs et plus de 250 étudiants-chercheurs (Postdoc, PhD et MSc). Il a aussi dirigé le groupe de recherche en Microélectronique et Microsystèmes de Polytechnique Montréal pendant plus de 15 ans et le département de génie électrique de cette même institution pendant 6 ans.

Abstract

Soft errors, induced by radiation, have a growing impact on the reliability of CMOS integrated circuits partly due to the progressive shrinking of device sizes. In this work, we explore a hierarchical methodology to model, analyze, and estimate Single Event Transients (SETs) propagation at different abstraction levels (gate and RTL levels). Gate level SET libraries are developed to abstract SET propagation conditions and probabilities from gate level models. At RTL level, these libraries are utilized to model the underlying probabilistic behavior of SET propagation as Probabilistic Automata (PA). Thereafter, the PAs of the RTL components are used to construct a Markov Decision Process (MDP) model of RTL designs. A probabilistic model checker is adapted to analyze the probability of SET propagation for all vulnerable nodes. Furthermore, a new estimation of the Soft Error Rate (SER) at RTL is developed. Experimental results demonstrate that our proposed framework is orders of magnitude faster than contemporary techniques and also can handle large designs such as 256-bit adders.

Our research also explores SRAM-based FPGAs that are increasingly popular in the aerospace industry for their field programmability and low cost. However, they suffer from cosmic radiation induced Single Event Upsets (SEUs), commonly known as soft errors. In safety-critical applications, the dependability of the design is a prime concern since failures may have catastrophic consequences. An early analysis of dependability of such safety-critical applications will enable designers to develop a design that meets the high availability and reliability requirements, such as the DO-254 standard. We explore methodologies based on probabilistic model checking, to analyze the dependability properties of safety-critical systems and to suggest required mitigation techniques, such as rescheduling, Triple Modular Redundancy (TMR) or TMR with less frequent scrubs for early design decisions. Starting from a high-level description of a system, a Markov (reward) model is constructed from the extracted Control Data Flow Graph (CDFG) and the failure/mitigation parameters for the targeted FPGA. Such an exhaustive model captures all the failures and repairs possible in the system within the radiation environment. Such early analysis is useful for designers of safety critical systems to assess the dependability and performability, saving both cost and effort.

We present some case studies to illustrate the applicability of the proposed approaches, and to demonstrate that a wide range of useful dependability and performability properties can be analyzed using our proposed methodology.