Enhanced Worst-case Timing Analysis of Ring-based Networks with Cyclic Dependencies using Network Calculus

Abstract

The recent research effort towards defining new communication solutions for cyber-physical systems (CPS), to guarantee high availability level with limited cabling costs and complexity, has renewed the interest in ring-based networks. This topology has been recently used for various networked cyber-physical systems (Net-CPS), e.g., avionics and automotive, with the implementation of many Real Time Ethernet (RTE) profiles. A relevant issue for such networks is to prove timing predictability, a key requirement for safety-critical systems. For the most common ring-based Real Time Ethernet (RTE) profiles, conducting such performance analyses has been greatly simplified due to their implemented time-triggered communication scheme, e.g. Master/slave or TDMA. Unlike these existing approaches, we are interested in this paper in event-triggered ring-based networks, which guarantee high resource utilization efficiency and (re)confi\-gura\-tion flexibility, at the cost of increasing the timing analysis complexity. The implementation of such a communication scheme on top of a ring topology actually induces cyclic dependencies, in comparison to time-triggered solutions. To cope with this arising issue of cyclic dependencies, only few techniques have been proposed in the literature, mainly based on Network Calculus framework, and consist in analyzing locally the delay upper bound in each crossed node, resulting in pessimistic end-to-end delay bounds. Hence, the main contribution in this paper is enhancing the delay bounds tightness of such networks, through an innovative global analysis based on Network Calculus, considering the flow serialization phenomena along the flow path. An extensive analysis of such a proposal is conducted herein regarding the accuracy of delay bounds and its impact on the system performance, i.e., scalability and resource-efficiency; and the results highlight its outperformance, in comparison to conventional methods.