Briefing
The core research problem in Multi-BFT systems is the performance bottleneck created by requiring a rigid global ordering of blocks produced by parallel consensus instances. This paper proposes Ladon, a novel Multi-BFT protocol that achieves Dynamic Global Ordering by assigning monotonic ranks to blocks based on their generation, which respects inter-block causality. This mechanism is achieved through pipelining the rank coordination with the consensus process and utilizing aggregate signatures to reduce overhead. The single most important implication is the elimination of blocking on slow instances, which fundamentally re-architects BFT systems to achieve unprecedented scalability and efficiency, demonstrating up to an 8x increase in peak throughput.
Context
The established theory of Byzantine Fault Tolerance (BFT) consensus, while providing strong security and finality, is fundamentally limited in scalability due to its high communication complexity, typically quadratic to the number of nodes. Multi-BFT was introduced to circumvent the single-leader bottleneck by running multiple leader-based consensus instances in parallel. However, this approach introduced a new, critical challenge → the necessity of deriving a total, global order for the output blocks across all parallel instances.
Prior designs addressed this by pre-assigning a global index before block creation, which inevitably led to performance degradation and instances blocking when a single parallel instance lagged. This limitation restricted the practical application of Multi-BFT to large-scale, high-throughput decentralized systems.
Analysis
Ladon’s core mechanism is the decoupling of block finalization from its position in the global sequence, enabling true parallel processing without synchronization delays. The protocol achieves this by replacing the static pre-assigned index with a Monotonic Rank that is dynamically coordinated across all parallel instances. When a block is proposed, it is assigned a rank, and the coordination of this rank is pipelined alongside the standard consensus voting process, minimizing protocol overhead.
This dynamic ranking ensures that blocks are globally ordered according to their generation, preserving the necessary causal dependencies between them. Furthermore, the system employs aggregate signatures combined with rank information to significantly reduce the overall message complexity of the global ordering process, a critical improvement over previous $O(n^2)$ or $O(n^3)$ BFT variants.
Parameters
- Peak Throughput Improvement → 8x – The approximate increase in peak throughput demonstrated by Ladon-PBFT over prior art in a Wide Area Network (WAN) setting.
- Latency Reduction → 62% – The approximate reduction in transaction latency achieved by the Ladon-PBFT implementation.
- Straggling Replicas Tested → 1 out of 128 – The fault tolerance scenario tested, demonstrating performance resilience even with a slow replica.
Outlook
This research opens a new avenue for scaling Byzantine Fault Tolerance by shifting the focus from simply parallelizing consensus to dynamically coordinating the global state. The immediate next step involves formalizing the security proofs for the dynamic rank coordination mechanism under various asynchronous network conditions. In the next 3-5 years, this theory could unlock the development of highly scalable, sharded blockchain architectures where cross-shard communication is coordinated with near-optimal latency, enabling a new class of high-performance decentralized applications that require massive transaction throughput without sacrificing BFT-level security guarantees.
