Performance Evaluation on Packet Transmission for Distributed Real-time Avionics Networks Using Forward End-to-End Delay Analysis

With growing avionics applications, the transmission of avionics data fl ows has been increasing in real-time avionics networks of aircraft. An Avionics Full DupleX switched Ethernet (AFDX), standardized as ARINC 664, is chosen as the backbone network for distributed real-time avionics systems as it o ff ers high throughput and does not require global clock synchronization. Estimating the end-to-end transmission delay to validate the network performance is essential for both certi fi cation and industrial research. Because of the various waiting times caused by the backlog (i.e., the pending packets in the output port of the visited switch), it is necessary and reasonable to compute the worst-case end-to-end transmission delay to validate network performance. Several approaches have been designed to compute the upper boundaries of end-to-end transmission delays, such as the Network Calculus approach and the Trajectory approach. In this paper, we focus on a new approach, Forward end-to-end delay Analysis (FA). This approach iteratively estimates the maximum backlog (i.e., number of pending packets) in each switch visited along the transmission path, so that the worst-case end-to-end transmission delay can be computed and the network performance evaluated. We also present the termination condition for this iterative estimation. The experiments demonstrate that this approach achieves a more accurate evaluation of transmission performance than the Network Calculus approach. A comparison with the exact upper boundaries obtained using the Model Checking approach shows the pessimism (i.e., overestimation) in FA. This paper analyses the reasons for that pessimism, and proposes future research.