Three-round consensus latency with high throughput and robustness

Determine whether a partially synchronous Byzantine fault tolerant (BFT) consensus protocol can achieve optimal total consensus latency of three rounds while simultaneously attaining high throughput and robustness; alternatively, prove a lower bound showing that any protocol meeting these throughput and robustness criteria must have total consensus latency of at least four rounds.

Background

The paper shows that Zaptos can achieve blockchain end-to-end latency equal to client–fullnode and fullnode–validator communication plus the consensus latency, provided execution and commit fit within one round. Thus, the overall latency becomes optimal when the underlying consensus latency is optimal.

The authors note that although the lower bound for consensus ordering latency under partial synchrony is three rounds, practical protocols that deliver high throughput and robustness generally incur at least an additional round for data dissemination, yielding a total consensus latency of four rounds. The open question asks whether the three-round total latency is attainable without sacrificing throughput and robustness, or whether a fundamental lower bound of four rounds exists for protocols with these properties.