Solana

Alpenglow: Solana’s New Consensus Protocol Built for Real-Time Blockchains

David Liang

Solana Ecosystem Lead

Published

July 10, 2025

Solana’s Agave client and consensus protocol are evolving to meet their stated roadmap goals to Increase Bandwidth and Reduce Latency. Alpenglow is Solana’s most significant consensus upgrade yet. A complete replacement of foundational components like TowerBFT and Proof-of-History ( @aeyakovenko’s original code), to achieve millisecond-level finality and Internet Capital Markets performance.

Alpenglow’s two core evolutions:

• Votor: a high-speed consensus protocol that finalizes slots in a single or double voting round, depending on validator responsiveness.
• Rotor: an optimized data propagation layer that uses stake-weighted relays and erasure coding to eliminate bandwidth bottlenecks.

The result: block finality in as little as 100-150 milliseconds.

Credit: Anza Labs

What’s Changing, And Why It Matters

TowerBFT and Proof-of-History helped Solana achieve unprecedented throughput and finality compared to earlier Smart Contract blockchains, such as 12 second finality, max TPS of 65,000. However, their design introduced bandwidth and latency bottlenecks for future network optimization and growth. Alpenglow replaces them entirely with Votor and Rotor.

Votor: Faster Finality with Fewer Rounds

Votor simplifies Solana’s consensus voting logic. It runs two voting modes concurrently:

• 80% fast path: If 80% of Solana stake votes, the block is finalized in a single round.
• 60% fallback path: If only 60% of stake is responsive, finality is achieved in two rounds.

Finalization occurs as soon as one of the paths completes. This dual-mode design allows Solana to finalize slots with significantly lower latency than before. This avoids probabilistic confirmations, long finality chains, and waiting for tower lockouts to resolve.

Rotor: Bandwidth-Optimal Block Propagation

Rotor replaces the old Turbine protocol for data distribution. Like Turbine, Rotor erasure-codes each block into smaller pieces or “shreds”, that can be reconstructed from a subset. Rotor introduces critical improvements:

• Single-layer relay structure reduces the number of data hops and propagation delays.
• Stake-weighted relay selection aligns validator incentives.
• Stateless repair and validation allow nodes to catch up to the tip of the chain quickly and securely.

The result: a data layer that scales efficiently and aligns bandwidth usage with stake distribution. This is crucial to support Solana’s globally distributed validator set without sacrificing performance.

Alpenglow’s Performance Breakthroughs

Alpenglow’s Performance:

• Median finality latency: ~150 ms
• Best-case finality: ~100 ms (under ideal network conditions)
• Resilience: Tolerates up to 20% adversarial stake and another 20% offline
• Throughput: Scales with available network bandwidth using Rotor’s optimized relay system

These new performance benchmarks make Solana faster, and now viable for latency-sensitive use cases. These include:

• Traditional equity or Central Limit Order Book trading systems
• Multiplayer games
• Live auctions
• Any other applications that previously required Web2-level bandwidth and latency guarantees.

For Validator Operators: Agave and the New Baseline

The Alpenglow upgrade is part of Agave, Solana’s validator client by Anza. Validator operators, especially those serving institutional digital asset clients, need to prepare for a higher performance bar:

• Latency matters. Finality happens in under 200 ms. Infrastructure must be tuned for relay speed and uptime under tighter time windows.
• Bandwidth is proportional to stake. With Rotor’s stake-weighted relay logic, validators must ensure their outbound networking capacity scales with their Total Active Stake.
• Client integration is not optional. Agave is the reference implementation for Alpenglow. Upgrading and validating infrastructure against Agave is essential to remain competitive.

Solana validator operators like Figment are ready to test and upgrade Agave. Institutional digital asset clients are asking for better transaction landing methods during periods of high congestion like January 2025. Performant infrastructure is required under sustained heavy network load. Solana wants to scale to be onchain NASDAQ. Increasing bandwidth with larger block limits of 100MM CUs, 1MM TPS and reducing latency with lower slot times. Agave’s Alpenglow upgrade delivers a path to onchain NASDAQ.

A New Class of Viable Applications

What happens when Solana finalizes slots in under 150ms?

• Faster onchain order books: Launch onchain Central Limit Order Books with equivalent order matching speed to a centralized exchange
• Payment Processors: streaming payments execute in near real time
• Video games: game state changes reflect instantly for all players

This is not theoretical. It’s measurable, repeatable, and already simulated with Solana’s mainnet validator and latency data.

Alpenglow moves Solana past “fast for a blockchain”, and into competition with the real-time systems powering today’s internet architecture.

Alpenglow Rollout Timeline

Alpenglow Announcement at Solana Accelerate Anza Lab’s Alpenglow Tweet

Per discussions with core devs and team at Agave, Alpenglow mainnet rollout is tentatively expected for Solana Mainnet in Q4 2025.

Figment’s Involvement

For institutional-focused Solana validator operators like Figment, Alpenglow is a critical infrastructure upgrade. Agave integrates Alpenglow’s consensus as their baseline implementation. Adoption is mandatory for Solana validator operators. Figment’s protocol, product and devops teams coordinate proactively integrate and test upgrades to Agave.rs.

Alpenglow isn’t just a protocol upgrade, but it is the clearest line between Solana’s bandwidth and latency performance. Only Possible On Solana is now: what we built pre-Apenglow and post-Alpenglow.

Learn More

The Alpenglow Whitepaper is available now. It includes:

• Full protocol specifications for Votor and Rotor
• Formal correctness proofs for safety and liveness
• Simulation results using Solana’s mainnet node and latency distribution
• Implementation notes for validator Solana validator operators and client developers

Solana’s consensus performance ceiling is raised. Now is the time for Figment to execute.

Sources:

Alpenglow: A New Consensus for Solana, by Anza Labs Alpenglow Whitepaper