Stochastic network calculus for analysis of latency on TTEthernet network

Time-triggered Ethernet (TTE) is a novel hybrid communication network with time-triggered and event-triggered messages. The determinacy of avionics full duplex (AFDX) switched Ethernet is improved by introducing the time-trigger (TT) flow. Even though TT flow has full time certainty, rate-constraint (RC) flow which is compatible with the virtual link (VL) in the AFDX network is still uncertain. Since the influence of fixed schedule slots of TT flow, traditional performance analysis methods used in the AFDX network are not suitable for the. In order to guarantee the real-time performance of RC flow, we propose two latency analysis models respectively based on the deterministic network calculus and stochastic network calculus. Under the deterministic network calculus, the latency upper bounds of RC flow are obtained by constructing the aggregate arrival curve of TT flow and service curve for RC flow. Under the stochastic network calculus, we compute the probabilistic upper bounds for RC flow by constructing two states Bernoulli distribution models for RC flow through Chernoff bound. Comparison results show that stochastic network calculus reduces the pessimism of deterministic network calculus on the performance analysis of RC flow. In addition, simulation results verify the correctness of two theoretical methods in some degree.