Poster: Optimizing Tree-based Quorum Certification for BFT Systems at Planetary-Scale

Recent research work on Byzantine fault-tolerant (BFT) protocols focuses on the question of how to improve scalability. A particular scalability-enhancing technique used in several BFT protocols such as ByzCoin, Motor, and Kauri is load balancing the communication through a tree-based dissemination and aggregation of quorum certificates. However, current evaluations of these algorithms often overlook how heterogeneous network latencies impact system performance. In this work, we first explore the influence of heterogeneous latencies as found in planetary-scale networks towards tree-based BFT protocol performance. Subsequently, we propose an optimization method which searches for an optimal communication tree by adapting tree configurations to given network conditions, thus minimizing system latency. Simulation-based experimentation reveals that the margin of latency improvements is substantial: An optimized configuration is up to 2.16× faster than the mean. We further discuss required modifications to support reconfigurations in Kauri with fast configurations, thus assisting automated deployments.