This is a segment of the LightSpeed Newsletter. Subscribe to read the full edition.
Solana Balidators have begun voting for SIMD-0326, the governance proposal for the new Alpenglow Consensus Protocol.
The proposal requires at least 33% validators to reach the quorum. As of this morning, at least 11.8% (135) of Solana’s approximately 1.3k validators are supported based on Dune.
Sending a transaction in Solana today takes around 12.8 seconds (32 slots * 0.4S slots time) to reach a critical finality, but sometimes it is informally called as having a soft/optimistic transaction finality of 500-600ms.
12.8 seconds is faster than the 12.8 minute finality of Ethereum’s Gasper Consensus, but pales in comparison to the thin speeds of 390-500ms, compared to the consensus speed of a new chain like Sui’s Mysticeti.
Alpenglow promises to collapse the 12.8-second time frame to an additional 100-150ms, bringing about a 100x improvement depending on where the validator is geographically located.
This does so by eliminating several core legacy pieces in Solana’s chain, such as Proof of History (POH), Tower BFT, and Gossip-based voting propagation.
The average Solana user tries to keep it simple, as they don’t know what these are and don’t have to worry about them.
POH and Tower BFT are twin core pieces of today’s Solana architecture, allowing validators to order blocks without interaction.
Think of POH as a pre-consensus encryption stopwatch that indicates that timestamps will block. These timestamps act as shared clocks where the Tower BFT algorithm helps validators for commitment. In this way, the Solana blockchain can continue to generate blocks without the need for a traditional chain and a typical synchronous consensus round.
This protocol design is useful for Solana, but there are trade-offs.
One creates the overhead cost of a voting transaction. This makes up something like 70% of Solana’s Onchain transactions (see the grey chart below). Voting transactions are clearly different from the type of transactions that a regular user sends to use the chain. They prove that the balleters have to pay to lock down the commitment to Tower BFT, or the costs of the business.
This has created many years of harmful problems for the Solana Balidators.
The larger the Solana Validator stock, the more revenue opportunities are given to produce the block. However, all validators pay the same fee to vote, regardless of the size of their investment.
This “fixed cost, variable income” economics causes large valiters to make large investments over time, effectively centralizing network interests to large variators because they are small. Network decentralization suffers because running validators is unprofitable unless they have deep pockets.
(It also explains why the Solana Foundation has an initiative to grant small validators in its first year of operation.)
Alpenglow aims to snag this issue by completely removing the voting fees. It comes with new components: voters and rotors.
Voters shift validator votes off-chain, but record tallies as compact on-chain certificates. Therefore, Alpenglow replaces slot-by-slot voting transactions with fixed fees and compresses the consensus into 1-2 rounds to achieve a target finality of 100-150ms.
Loading Tweets…
