FnF-BFT: Exploring Performance Limits of BFT Protocols

We introduce FnF-BFT, a parallel-leader byzantine fault-tolerant state-machine replication protocol for the partially synchronous model with theoretical performance bounds during synchrony. By allowing all replicas to act as leaders and propose requests independently, FnF-BFT parallelizes the execution of requests. Leader parallelization distributes the load over the entire network -- increasing throughput by overcoming the single-leader bottleneck. We further use historical data to ensure that well-performing replicas are in command. FnF-BFT's communication complexity is linear in the number of replicas during synchrony and thus competitive with state-of-the-art protocols. Finally, with FnF-BFT, we introduce a BFT protocol with performance guarantees in stable network conditions under truly byzantine attacks. A prototype implementation of prot outperforms (state-of-the-art) HotStuff's throughput, especially as replicas increase, showcasing prot's significantly improved scaling capabilities.