Automatic performance tracking of a SpaceWire network

Verification of complex networks, especially meshed networks created of routers, can become quite difficult. There are several parameters influencing the actual data throughput, e.g., congestion in the network, transmission rates at the inputs of the network or between routers as well as the reception rate of data. An appropriate model is required to evaluate the network performance. This model can be defined at different levels of detail whereas the more detailed levels are considered to be more precise with respect to the real hardware behavior. The objective of this paper is to define such a model and to provide random constraint stimuli for an automatic tracking of the network performance. The model consists of a meshed network with a fixed topology using HDL descriptions of the SpaceWire routers in our system. However, this approach can also be applied to other network topologies. Precisely mimicking input and output data streams that are applied to the network is crucial to identify inefficient data paths. These data streams are generated dependent on predefined high level constraints (e.g. packet lengths, transmission rates) that are transferred into lower level constraints to finally create the required stimulus. The quality of the system, dependent on a specific data stream, is determined among others by tracking and analyzing the transition time of packets from source to destination.