Fully-autonomous cyber-physical systems (CPS) such as autonomous vehicles, drones, and robots today are not engineered as rigorously as aircraft, and thus are not as reliable. In fact, these CPS will likely experience more failures in the future, since they will have a cumulative operation time several orders of magnitude more than that of airplanes. Thus, it is essential that we bring the reliability of today’s commercial aircraft systems (“ultra-reliability”) to the next generation of fully-autonomous CPS. One of the main challenges that needs to be addressed in this regard is ensuring high reliability with minimal cost in the presence of environmental disturbance sources such as electro-magnetic interference and thermal effects. To this end, I will present the first provably safe reliability analysis of Ethernet-based distributed real-time systems in the presence of environmentally-induced Byzantine errors. The analysis helps quantify and systematically evaluate the reliability tradeoffs involved when using different replication schemes. In this talk, I will discuss the key features of our analysis, including how we tackle reliability anomalies and how we quantify a CPS application’s inherent robustness to occasional failures, and my plans for future research.