page d'accueil   sommaire   institut  

Parallel Architectures for real-Time Systems [Parallel Architectures for Real-Time Systems] (PARTS)
Instituts et centres de recherche / institutes and research centers - Liste des Centres (unité ULB709)

Les systèmes temps réel sont des systèmes pour lesquels le bon fonctionnement dépend non seulement de la correction des résultats de calcul, mais aussi des instants où ces résultats sont produits (typiquement avant une échéance stricte fixée au préalable). De tels systèmes sont couramment utilisés dans un grand nombre d'applications industrielles: l'automobile, l'avionique, le contrôle de processus industriel et le grand public : la communication, les systèmes de calcul portables et des systèmes embarqués en général etc.Aujourd'hui, grâce aux techniques d'intégration des circuits électroniques très denses (l'intégration en 3 dimensions par exemple), il est possible de produire des systèmes contenant plusieurs dizaines d'unités de traitement et de mémoire (systèmes parallèles : Multi-Processor System-on-Chip). De telles plateformes permettent donc l'implémentation des applications complexes ayant une grande demande en capacité de calcul. Le défi est alors de définir dans la couche logicielle (le système d'exploitation), des techniques efficaces et réalistes de gestion d'une telle quantité de ressources (le problème d'ordonnancement - scheduling).Les techniques d'ordonnancement existantes couvrent principalement les systèmes mono processeur et ne peuvent pas être appliquées tels quelles aux systèmes parallèles. De plus, ces techniques ne tiennent pas comptes des réalités physiques d'implémentation des circuits intégrés, ni des contraintes supplémentaires introduits par de tels systèmes : l'optimisation de la surface occupée (i.e. le coût), la puissance dissipée (l'autonomie du système), les aspects thermiques (la vitesse de fonctionnement maximale pour une partie de système), etc.Dès lors, il est impératif de réaliser cette recherche dans une approche co-design, qui consiste à combiner les aspects matériels et logiciels en parallèle. [Real-time systems are defined as those systems in which the correctness of the system depends not only on the logical result of computations, but also on the time at which the results are produced. Real-time computing systems are widely used in many industrial applications. Examples of application domains that require real-time computing include: Control of engines, Chemical and nuclear plant control, Traffic, Time-critical packet communications, Flight control systems, Military systems, Space missions, Virtual reality, Railway switching systems, and Robotics.A major misconception about real-time systems is to think that they are equivalent to fast systems. Of course, minimizing the computation duration is helpful in satisfying the timing constraints, but it is not enough to meet all hard timing constraints. Instead of ensuring fast computation, in real-time systems we are concerned with a most important principle, called the predictability, i.e., the ability to predict, a priori, whether the system can meet all hard (also termed critical) timing requirements.]



coordonnées / contact details


Parallel Architectures for real-Time Systems [Parallel Architectures for Real-Time Systems]
tel +32-2-650.55.88 / 30.60, fax +32-2-650.56.09 / 24.82, parts@ulb.ac.be
Campus de la Plaine, Bat. NO, 8ème
CP212, boulevard du Triomphe, 1050 Bruxelles



responsables / head


Prof. Joël GOOSSENS Dragomir MILOJEVIC


composition / members


Yannick ALLARD Olivier DESENFANS Antonio PAOLILLO Benigno RODRIGUEZ LOBERA Vladimir SVOBODA Mohamed Amine YOUSSEF


projets / projects


Ordonnancement temps réel et contraint par l'énergie [Power-Aware real-time scheduling]
Ordonnancement temps réel et contraint par l'énergie [Energy consumption and battery lifetime are nowadays major constraints in the design of mobile embedded systems. Amongst all hardware and software techniques aimed at reducing energy consumption, supply voltage reduction, and hence reduction of CPU speed, is particularly effective. This is because CPU requires a large amount of energy (e.g., 30W at maximal frequency for an Intel P4 Mobile 1.8GHz[1]) and the energy consumption of the processor is usually at least quadratic in the speed of the processor (see [1] for more details). The aim is thus to minimize the processor frequency as much as possible while satisfying the performance constraints of the system. Many power-constrained embedded systems are built upon multiprocessor platforms because of high-computational requirements and because multiprocessing often significantly simplifies the design. As pointed out in [2] and [3], another advantage is that multiprocessor systems are theoritically more energy efficient than equally powerful uniprocessor platforms because raising the frequency of a single processor results in a multiplicative increase of the consumption while adding processors leads to an additive increase. We address the problem of determining one or several processor speeds which involve significant power savings while system is running. The determined speeds must satisfy all the temporal constraints of the system. In the second part of our research, we investigate the various models of battery and include their behavior into the scheduling algorithms in order to take into account a more realist supply voltage delivery.]

Services matériels pour des systèmes multi processeur temps réel [Hardware services for MPSoC with real-time operating systems]
Services matériels pour des systèmes multi processeur temps réel [In recent years we have witnessed a paradigm shift in computer systems. Increasing the frequency has given way to multi-core architectures exploiting the parallelism. In the field of embedded systems, such a vision is seen in the form of Multi-Processor System-on-Chip - MPSoC. The advantages of such a platform in comparison with a uni-processor one are multiples in several domains like power consumption, scalability and reusability. In the same time, a lot of existing systems need Real Time Operating Systems not only to guarantee a given treatment capacity but also to guarantee a deadline for multiple tasks. To break with the sub-optimality of the actual philosophy consisting to see the design of software and hardware as two worlds apart, the aim of this thesis is, through a co-design methodology, to design a configurable MPSoC environment with services tailored directly for high level real-time scheduling algorithms.]



publications





theses


Geoffrey Nelissen, Efficient Optimal Multiprocessor Scheduling Algorithms for Real-Time Systems, 2013

Vincent Nelis, Energy-Aware Real-Time Scheduling in Embedded Multiprocessor Systems, 2011

Vandy Berten, Stochastic Approach to Brokering Heuristics for Computational Grids, 2007

Joël Goossens, Scheduling of Hard Real-Time Periodic Systems with Various Kinds of Deadline and Offset Constraints, 1999



collaborations


Sanjoy Baruah, University of North Carolina at Chapel Hill, Computer Science department, Chapel Hill, Etats-Unis (USA)

Pascal Richard, Université de Poitiers, Poitiers, France

Shelby Funk, University of Georgia, Computer Science Department, Athens, Etats-Unis (USA)

Nathan Fisher, Wayne State University, Computer Science Department, Detroit, Etats-Unis (USA)



mots clés pour non-spécialistes / keywords for non-specialists


co-design multi-processor system-on-chip ordonnancement systèmes embarqués systèmes temps réel


disciplines et mots clés / disciplines and keywords


Sciences de l'ingénieur

co-design multi-processor system-on-chip ordonnancement temps réel


codes technologiques DGTRE


Électronique Informatique, théorie des systèmes Sciences de l'ordinateur, analyse numérique, systèmes, contrôle