Fault injection and reliability analysis on time sensitive networking

Abstract

Time Sensitive Networking (TSN) provides high performance communication using time scheduling to achieve deterministic latency and jitter. In theory, the rigor of a TSN schedule is the key to achieve deterministic communication. However, real devices are prone to errors. TSN-based applications require both fault-tolerance and end-to-end latency guarantee. In this work we identify the limitations of the TSN scheduling method, enhancing the simulation model NeSTiNg to enable fault-injection. The fault-injection is performed by introducing new components, and modules to emulate permanent, transient and intermittent faults based on probability distributions. We evaluate the effectiveness of the fault-injection comparing the latency and jitter results to the ones expected by the schedule generator TSNsched, also presenting a technique to increase network reliability on faulty scenarios. Moreover, we analyze the the schedule generated by TSNsched using criteria such as path and transmission window consistency. In conclusion, we were able to prove the effectiveness of the fault-tolerant techniques found on the state-of-the-art, compiling all the results and impacts on scheduled and non-critical traffic.