Master Thesis

Fast Paths in BFT Consensus: A Comparative Framework

Fast finalization is a recurring goal in Byzantine fault-tolerant (BFT) consensus: when conditions are favorable, a protocol may try to commit in fewer rounds than its standard execution. One way to achieve this is through a dedicated fast path, that is, an optimistic execution path that succeeds only under stronger assumptions or with extra resources. FaB Paxos [1] initiated this line of work, and later results refined its resilience bounds and clarified which optimistic latencies are achievable in closely related broadcast settings [2, 3]. More recent protocols such as Banyan [4] and Kudzu [5] revisit these ideas in modern BFT protocol designs.

This thesis asks the question: under which assumptions does a fast path exist in BFT consensus, and what does it cost? The student will compare a selected corpus of protocols in a common framework, focusing on dimensions such as synchrony assumptions, fault thresholds, quorum structure, replica count, and optimistic latency. A central part of the work will be to reconcile terminology and assumptions across papers so that the comparison is precise rather than merely descriptive.

The expected outcome is a structured taxonomy of fast-path designs and their trade-offs. The thesis should clarify which costs appear to be fundamental, which are tied to specific models, and where recent protocols genuinely improve over classical constructions. A strong thesis may also identify open gaps or mismatches in the literature, but the primary goal is a structured comparative framework that makes the design space easier to understand.

References

[1] Jean-Philippe Martin and Lorenzo Alvisi. Fast Byzantine Consensus. IEEE Transactions on Dependable and Secure Computing, 3(3):202-215, 2006.

[2] Anton Kuznetsov, Daniil Tonkikh, and Yanhong A. Zhang. Revisiting Optimal Resilience of Fast Byzantine Consensus. ACM PODC 2021.

[3] Ittai Abraham, Kartik Nayak, Ling Ren, and Zhuolun Xiang. Good-case Latency of Byzantine Broadcast: a Complete Categorization. ACM PODC 2021.

[4] Yann Vonlanthen, Jakub Sliwinski, Massimo Albarello, and Roger Wattenhofer. Banyan: Fast Rotating Leader BFT. 25th International Middleware Conference, 2024.

[5] Victor Shoup, Jakub Sliwinski, and Yann Vonlanthen. Kudzu: Fast and Simple High-Throughput BFT. 39th International Symposium on Distributed Computing (DISC 2025). (talk from Victor Shoup)

Contact Ulysse Pavloff and Juan Villacis for more information.

Nature of the project: Theory 80%, Systems 20%.