Scalable and Probabilistic Leaderless BFT Consensus through Metastability (1906.08936v2)

Abstract: This paper introduces a family of leaderless Byzantine fault tolerance protocols, built around a metastable mechanism via network subsampling. These protocols provide a strong probabilistic safety guarantee in the presence of Byzantine adversaries while their concurrent and leaderless nature enables them to achieve high throughput and scalability. Unlike blockchains that rely on proof-of-work, they are quiescent and green. Unlike traditional consensus protocols where one or more nodes typically process linear bits in the number of total nodes per decision, no node processes more than logarithmic bits. It does not require accurate knowledge of all participants and exposes new possible tradeoffs and improvements in safety and liveness for building consensus protocols. The paper describes the Snow protocol family, analyzes its guarantees, and describes how it can be used to construct the core of an internet-scale electronic payment system called Avalanche, which is evaluated in a large scale deployment. Experiments demonstrate that the system can achieve high throughput (3400 tps), provide low confirmation latency (1.35 sec), and scale well compared to existing systems that deliver similar functionality. For our implementation and setup, the bottleneck of the system is in transaction verification.

• The paper’s main contribution is the Snow family of protocols that use network subsampling and metastability to enable leaderless BFT consensus.
• It demonstrates high performance with up to 3400 transactions per second and low latency, offering probabilistic safety even under adversarial conditions.
• By eliminating PoW and leveraging efficient network sampling, the protocols achieve robust Byzantine fault tolerance and energy efficiency for decentralized applications.

Scalable and Probabilistic Leaderless BFT Consensus through Metastability

The paper introduces a suite of protocols for achieving leaderless Byzantine fault-tolerant (BFT) consensus named the Snow family, which is engineered around a metastable mechanism facilitated through network subsampling. The primary aim of this work is to address scalability and throughput limitations inherent in traditional and Nakamoto-style consensus methods by leveraging probabilistic safety guarantees, allowing for a highly scalable and efficient consensus mechanism. The key novelty lies in its concurrent, leaderless operation which promises high throughput, in contrast to conventional approaches that tend to falter with large-scale deployment owing to intricate communication overhead.

Core Contributions and Methodology

The source of this innovation stems from the realization that traditional consensus protocols are hindered by either quadratic communication complexity or significant energy expenditure due to their reliance on proof-of-work (PoW) as seen in Nakamoto protocols. The proposed Snow family protocols operate instead through metastability, where network nodes sample random subsets to steer decisions towards consensus states. This family comprises protocols named Slush, Snowflake, Snowball, and an advanced multi-decree protocol named Avalanche, all designed to escalate network dynamics towards a definite decision state leveraging probabilistic guarantees.

• Slush serves as a foundational protocol introducing metastability through network sampling without immediate Byzantine resistance, setting the stage for further enhancements.
• Snowflake incorporates resilience to Byzantine faults with a counter mechanism ensuring decision persistence under adversarial settings.
• Snowball extends Snowflake by introducing confidence counters to mitigate transient sampling variations, leading to increased robustness.
• Avalanche, constructed atop the Snow family, implements a Directed Acyclic Graph (DAG) structure to foster high transaction throughput and rapid consensus across potentially conflicting transactions without necessitating PoW.

Performance and Results

The empirical evaluations presented in the paper highlight Avalanche's potential to sustain high throughput, achieving transaction rates of up to 3400 transactions per second (tps). This performance metric significantly outstrips that of Bitcoin and aligns competitively with modern consensus systems like Algorand, yet with markedly superior latency metrics (1.35 seconds average confirmation time across 2000 nodes). Avalanche accomplishes this while maintaining energy efficiency by eschewing the energy-intensive PoW component intrinsic to Nakamoto protocols.

Theoretical Guarantees and Implications

From a theoretical perspective, the adoption of metastability ensures that the consensus can be achieved under probabilistic safety guarantees even under significant adversarial presence. This paper characterizes the conditions under which metastability delivers these guarantees and discusses the ramifications of sub-sampling accuracy on consensus failure probabilities, which can be configured to be exceedingly low (below $10^{-9}$).

The shift towards leaderless consensus in such a manner introduces new paradigms for internet-scale consensus, distinguishing itself profoundly from classical quorum-based BFT and extends its utility into permissionless systems. This has critical implications for decentralized financial applications and blockchain systems where the cost of consensus decisions previously constrained system scalability.

Future Directions

The leaderless and energy-efficient design of these protocols postulates transformative impacts on blockchain technology and distributed consensus. Future exploration may include deploying these protocols in heterogeneous and dynamically scaling environments, assessing live network challenges like churn more closely, and integrating this scalable consensus foundation with modern decentralized applications (dApps). Additionally, understanding real-world adversarial models built around network timing and mode suggests an enticing avenue for further security analysis and protocol fortification.

In conclusion, the paper's Snow family of consensus protocols, culminating in Avalanche, efficiently balance scalability, decentralization, and low energy usage, presenting a significant step forward in consensus design that aligns keenly with the trajectory of distributed ledger technologies.