Stability of time-sensitive networks: strategies for partial regulation

In time-sensitive networks, regulators can be used to reshape traffic, and their usage may be necessary to guarantee stability by providing worst-case delay bounds. In this project, I study partial regulation of time-sensitive networks with cyclic dependencies, focusing on performance analysis. Using recently developed Network Calculus tools, such as TFA (Total Flow Analysis), TFA++, FP-TFA (Fixed Point Total Flow Analysis) I study how delay bounds can be improved by varying position and number of per-flow regulators (PFRs) in a FIFO-per-class network. The problem is non-trivial because in very complex networks the number of possible combinations of positions becomes huge. I find that the choice of position and number of regulators can bring to significant improvements in the computed delay bounds. The main part of the analysis is done on a synthetic use-case network. The results of the analysis are discussed and some heuristics for the general case are inferred. I apply the work done to a realistic network by proposing an algorithm that aims to jointly find good positions and number of regulators in a general topology through a parametric analysis. In the realistic case, I find significant correspondences with the synthetic use-case results and considerations.