Hard real-time systems, such as the one found in the field of automotive, avionics or distribution of energy, are nowadays developed in the context of distributed architectures. In such systems, data streams have end-to-end network constraints to fulfill between their production and their consumption. The verification of the proper sizing of network resources compared to application needs is a normative requirement given the certification objectives of this type of systems.

Hard real-time systems, such as the one found in the field of automotive, avionics or distribution of energy, are nowadays developed in the context of distributed architectures. In such systems, data streams have end-to-end network constraints to fulfill between their production and their consumption. The verification of the proper sizing of network resources compared to application needs is a normative requirement given the certification objectives of this type of systems. Existing approaches to design distributed hard real-time system, for example based on the use of an arbitration policy to access to a bus-based network or when using network switches with full-duplex links, requires the oversizing of network resources. In this PhD thesis, we propose to model and characterize a new paradigm to propagate real-time data, and not simply of frames, within a multi-switch architecture in order to increase the utilization made of the network capacities and thus reduces the costs of such system. In particular, the definition and the study of data propagation & aggregation strategies will be carried out taking into account the end-to-end constraints of these data. Validation will be carried out by simulation and possibility through experimental evaluations.