Alpenglow: Solana New Consensus Paradigm

Under the new architecture, the efficiency of transaction finality will increase by 100 times.

Written by: Pzai, Foresight News

On the evening of May 19, Anza, the developer studio previously split from Solana Labs, released the new consensus layer protocol Alpenglow for Solana. This protocol changes the TowerBFT and PoH consensus mechanisms, employing a new component called Votor for voting and block finality, while using the Rotor component to improve Solana's existing block propagation protocol. It is built on top of Turbine (the Solana version of sharding) and optimizes bandwidth usage by utilizing a single-layer relay node based on stake.

Roger Wattenhover, the research director of Anza, stated at Solana Accelerate that the new consensus mechanism will significantly reduce the current transaction finality duration (12.8s) to 150 ms. In terms of the development process, Alpenglow has completed prototype testing and is expected to deploy a testnet around mid-2025, followed by a mainnet deployment later in 2025 after the Solana Improvement Document (SIMD) proposal is approved. Compared to the current Solana mainnet, Alpenglow simplifies the architecture and optimizes data propagation efficiency, bringing its performance closer to traditional internet infrastructure, making it suitable for scenarios such as high-frequency trading and real-time payments. This article gives you an overview of Alpenglow, known as the "Solana Consensus Restructuring."

Votor will handle the consensus logic and replace TowerBFT. It does not rely on the current node's "gossip" model, but instead votes on block finality through "direct communication." As a core component of the Alpenglow protocol, Votor's core innovations are reflected in its communication patterns, voting mechanisms, and performance optimizations.

First, Votor does not rely on the current node's "gossip" model, using direct peer-to-peer communication and dynamic grouping strategies (divided by equity weight or geographical location), significantly reducing redundant message transmission and lowering network latency.

Secondly, Votor introduces a tiered equity voting mechanism: if the first round of block receives over 80% equity support, notarization is completed directly; if the support rate is between 60% and 80%, a second round of rapid confirmation is initiated through a parallel voting track, while allowing nodes to actively skip voting when detecting block delays or risks, in order to avoid resource waste. From the data, when the overall validator threshold is below 60%, the delay can be controlled to around 100 ms.

Rotor focuses on improving block propagation efficiency and network resource allocation. By integrating Turbine sharding technology, it enhances the existing block propagation protocol of Solana. In practice, Rotor replaces the traditional multi-layer relay model with a single-layer relay node architecture, splitting block data into lightweight shards and dynamically optimizing transmission paths, significantly reducing network complexity and transmission latency.

In addition, Rotor introduces an adaptive propagation algorithm that monitors network status in real-time and switches congested paths. By combining lightweight data verification to reduce computational overhead, it significantly enhances propagation speed and fault tolerance. In terms of performance, Rotor compresses block propagation latency to the millisecond level, supporting Solana in achieving a high throughput target of 50,000 TPS, meeting the needs of high-frequency scenarios such as DeFi liquidation and real-time payments.

Overall, the Alpenglow protocol reduces operational risks across the entire chain and simplifies architecture by removing the PoH mechanism. It replaces Tower BFT consensus with Votor while adopting a stake-driven 1-2 round voting process, achieving block finality within 100-150 milliseconds without relying on optimistic confirmation. Rotor optimizes Turbine sharding with a single-layer relay system, enhancing propagation efficiency to the physical network latency limit through dynamic global bandwidth optimization and adaptive path selection, leaving the primary bottleneck as the underlying network transmission speed. At the same time, the system's resilience is significantly enhanced, capable of withstanding extreme scenarios of 20% malicious nodes and 20% staked offline, improving attack resistance and fault tolerance. Ultimately, Alpenglow compresses transaction finality to the millisecond level, providing underlying support for high-frequency trading, real-time payments, and large-scale on-chain applications.