Shoal++: High Throughput DAG BFT Can Be Fast!

TL;DR

Shoal++ presents a DAG-based partially synchronous BFT that narrows the latency gap to optimal PBFT-like protocols while preserving DAG throughput. It achieves about $4.5md$ end-to-end latency in the fault-free case by integrating three mechanisms: a Fast Direct Commit for anchors, dynamic re-interpretation of anchor schedules to increase anchor frequency, and running multiple DAGs in parallel to reduce queuing. Empirical evaluation shows Shoal++ matching or surpassing state-of-the-art DAG-BFT throughput with substantially lower latency, and robust performance under faults and network drops. The work highlights the value of parallel DAGs and adaptive anchor management for scalable, low-latency BFT in geo-distributed settings.

Abstract

Today's practical partially synchronous Byzantine Fault Tolerant (BFT) consensus protocols trade off low latency and high throughput. On the one end, traditional BFT protocols such as PBFT and its derivatives optimize for latency. They require, in fault-free executions, only 3 message exchanges to commit, the optimum for BFT consensus. However, this class of protocols typically relies on a single leader, hampering throughput scalability. On the other end, a new class of so-called DAG-BFT protocols demonstrates how to achieve highly scalable throughput by separating data dissemination from consensus, and using every replica as proposer. Unfortunately, existing DAG-BFT protocols pay a steep latency premium, requiring on average 10.5 message exchanges to commit a transactions. This work aims to soften this tension and proposes Shoal++, a novel DAG-based BFT consensus system that offers the throughput of DAGs while reducing commit latency to an average of 4.5 message exchanges. Our empirical findings are encouraging, showing that Shoal++ achieves throughput comparable to state-of-the-art DAG BFT solutions while reducing latency by up to 60%.

Figures (8)

• Figure 1: Narwhal's round-based DAG from a validator's local point of view.
• Figure 2: Bullshark Commit example. Replica 4 observes $f+1 = 2$ votes for anchor A1, and directly commits it. Replica 1 observes only one vote ($< f+1$), and cannot directly commit A1; it commits A1 indirectly upon committing A2 and its causal history, ordering A1 before A2.
• Figure 3: Bullsharks's direct commit rule requires f+1 certified vote nodes, while Shoal++'s direct commit rule requires $2f+1$ (uncertified) node proposals.
• Figure 4: An illustration with four anchors in round $1$, using the pre-assigned order: blue, red, green, pink. The blue anchor commits directly; it has $\geq f+1$ references from nodes in round $2$. The red anchor is only indirectly committed upon directly committing the yellow anchor in round $4$. The green and pink anchors must wait to be ordered until the red anchor is resolved.
• Figure 5: No failures with 100 replicas.

Theorems & Definitions (5)

• Lemma 1
• proof
• Lemma 2
• proof
• Theorem 1