Latency-Optimal View Synchronization for 2f+1 BFT Consensus in the Presence of Faults
D. Lihotský (TU Delft - Electrical Engineering, Mathematics and Computer Science)
Jérémie Decouchant – Mentor (TU Delft - Electrical Engineering, Mathematics and Computer Science)
H.J. Griffioen – Graduation committee member (TU Delft - Electrical Engineering, Mathematics and Computer Science)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
Byzantine fault-tolerant (BFT) consensus underpins modern state machine replication and blockchain systems. Recent consensus protocols reduce the classical 3f+1 replica requirement to 2f+1 by leveraging trusted execution environments, but view synchronization in this setting has received little systematic attention and typically relies on exponential timeout backoff, which has unbounded worst-case latency. On the other hand, in the 3f+1 setting, recent view synchronization work has focused on minimizing asymptotic message complexity, with limited consideration of the latency cost this imposes. This work addresses both gaps. We first adapt the message-optimal LP22 and Lumiere synchronizers to the 2f+1 setting. We then design two latency-optimized synchronizers tailored to the 2f+1 setting: a broadcast-based protocol and a leader-based protocol, which we call Babette. We implemented all four algorithms on top of the OneShot consensus protocol and evaluated them on the DAS-5 cluster across network sizes up to n=51 and fault scenarios ranging from no failures to the maximum tolerated f. With f failures, Babette achieves around 5.5× higher throughput and correspondingly lower latency than Lumiere across all tested network sizes, demonstrating that asymptotic message complexity-optimality does not immediately translate to performance improvement when timeout constants dominate end-to-end performance.