Performance analysis of FlexRay-based systems using real-time calculus, revisited

The FlexRay protocol [4] is likely to be the de facto standard for automotive communication systems. Hence, there is a need to provide hard performance guarantees on properties like worst case response times of messages, their buffer requirements, end-to-end latency (for example, from sensor to actuator), etc., for FlexRay based systems. The paper [11] provides an analysis for finding worst case response times of the messages transmitted on the FlexRay bus, but the analysis is done using ILP formulation and is thus computationally expensive. The paper [5] models the FlexRay in the analytic framework of Real-Time Calculus [12, 3] and is compositional as well as scalable. In this paper, we show that the analysis of [5] may lead to results that are over optimistic; in particular, we show that obtaining the "upper service curves" is not trivial and does not follow the reasoning of the "lower service curves" which the authors obtain. We also provide tighter "lower service curves" than that of [5]. Finally we show that our model allows the messages to be of variable size which is not the case with [5].