Paper 2024/472

Sailfish: Towards Improving Latency of DAG-based BFT

Abstract

The traditional leader-based BFT protocols often lead to unbalanced work distribution among participating parties, with a single leader carrying out the majority of the tasks. Recently, Directed Acyclic Graph (DAG) based BFT protocols have emerged as a solution to balance consensus efforts across parties, typically resulting in higher throughput compared to traditional protocols. However, existing DAG-based BFT protocols exhibit long latency to commit decisions. The primary reason for such a long latency is having a leader every 2 or more ``rounds''. Even under honest leaders, these protocols require two or more reliable broadcast (RBC) instances to commit the proposal submitted by the leader (leader vertex), and additional RBCs to commit other proposals (non-leader vertices). In this work, we present \name, the first DAG-based BFT that supports a leader vertex in each round. Under honest leaders, \name maintains a commit latency of one RBC round plus $1\delta$ to commit the leader vertex (where $\delta$ is the actual transmission latency of a message) and only an additional RBC round to commit non-leader vertices. Furthermore, we extend \name to \multiname, which facilitates multiple leaders within a single round and commits all leader vertices in a round with a latency of one RBC round plus $1\delta$. Through experimental evaluation, we demonstrate that our protocols achieve significantly better latency compared to state-of-the-art DAG-based protocols, with slightly better throughput.