End-to-end latency and temporal consistency analysis in networked real-time systems

Critical embedded systems are often designed as a set of real-time tasks, running on shared computing modules, and communicating through networks. Because of their critical nature, such systems have to meet strict timing properties. To help the designers to prove the correctness of their system, the real-time systems community has developed numerous approaches for analysing the worst case scenarios either on the processors (e.g., worst case response time of a task) or on the networks (e.g., worst case traversal time of a message). These approaches provide results only for local components behaviours. However, there is a growing need for having a global view of the system, in order to determine end-to-end properties. Such a property applies to functional chains which describe the behaviour of sequences of tasks. We propose an approach to analyse worst case behaviour along functional chains in critical embedded systems. It is based on mixed integer linear programming (MILP) and is general in the sense that it can be applied to a variety of end-to-end properties. This paper focuses on two essential properties: end-to-end latency and temporal consistency. This work was supported by the French National Research Agency within the SATRIMMAP project.