Scale or Die 2025: A Practical Approach to Multi-Concurrent Proposers

In a groundbreaking presentation at the Solana Conference 2025, Patrick O'Grady, founder of Commonware, unveiled a revolutionary approach to implementing multi-concurrent proposers on the Solana blockchain. This innovative solution promises to enhance Solana's scalability and decentralization while addressing critical challenges in blockchain design.

Summary

Patrick O'Grady, formerly the head of engineering at Apple Labs and a veteran of Coinbase's crypto engineering team, introduced Commonware's novel approach to implementing multi-concurrent proposers on Solana. The presentation focused on a simpler mechanism design that could potentially solve the challenges outlined in Anatoly Yakovenko and Max Levchin's recent blog post on the importance of multi-concurrent proposers for decentralized exchanges and markets.

O'Grady's solution, termed "Sequencer Driven Decoupled State Machine Replication," aims to address the issues of duplicate transactions and fund exhaustion that can arise in multi-concurrent proposer systems. By pre-selling network capacity to out-of-protocol sequencers and allocating accounts to specific sequencers, this approach simplifies the implementation of multi-concurrent proposers while maintaining the desired properties of censorship resistance and fair ordering.

The proposed system offers significant benefits for applications and users, including the ability for sequencers to determine transaction ordering and fee structures for their allocated accounts. This approach could potentially revolutionize how blockchain capacity is managed and how applications interact with their users on the Solana network.

Key Points:

The Need for Multi-Concurrent Proposers

Multi-concurrent proposers are crucial for decentralized exchanges and markets on blockchain networks. They allow for the running of auctions and prioritization of cancels, which is essential for protecting market makers from potential exploitation between batches. The current batched nature of blockchains can leave market makers vulnerable to having their positions exploited by those engaging in MEV (Miner Extractable Value) activities.

Anatoly Yakovenko and Max Levchin's blog post highlighted the importance of multi-liter protection, where multiple validators concurrently propose blocks or chunks with transactions. This approach ensures that market makers have the option to cancel their orders and get ahead of potential adverse trades.

Decoupled State Machine Replication (DSMR)

O'Grady introduced the concept of Decoupled State Machine Replication (DSMR) as a foundation for his proposed solution. Unlike traditional blockchain designs where block building, replication, and verification occur in a strict sequence, DSMR allows for these processes to happen concurrently across all validators.

This approach potentially offers higher throughput by leveraging parallel processing, although it introduces some additional latency in block settlement. The key innovation in DSMR is the decoupling of each step in the state machine replication process, allowing for greater flexibility and potential scalability.

Challenges in Multi-Concurrent Proposer Systems

O'Grady highlighted two primary challenges that can arise in multi-concurrent proposer systems: duplicate transactions and fund exhaustion. Duplicate transactions occur when multiple validators include the same transactions in their proposed blocks, leading to wasted throughput. Fund exhaustion happens when an account's balance is depleted in the first transaction, rendering subsequent transactions from that account invalid due to insufficient fees.

These issues can significantly reduce the effective throughput of the network and open up opportunities for malicious actors to exploit the system. The complexity of resolving these challenges in an asynchronous execution environment makes implementing multi-concurrent proposers particularly difficult.

Sequencer Driven Decoupled State Machine Replication

To address these challenges, O'Grady introduced Commonware's solution: Sequencer Driven Decoupled State Machine Replication. This approach pre-sells network capacity to a set of out-of-protocol sequencers for each epoch. These sequencers are then responsible for allocating capacity to accounts under their management.

Key features of this system include:

1. Accounts are allocated to exactly one sequencer at a time, preventing duplicate transaction submissions.
2. Sequencers maintain the fee-paying relationship with the network, simplifying fee management.
3. Applications can act as sequencers, giving them control over transaction ordering and fee structures for their users.

This design effectively outsources the complexity of managing multi-concurrent proposers to the sequencers while maintaining the desired properties of censorship resistance and fair ordering.

Implementation and Benefits

The implementation of this system involves limiting valid blocks to those that include references from a sufficient number of sequencers (M of N committee certificates). Users and applications can register with multiple sequencers to ensure censorship resistance.

This approach offers several benefits:

1. Simplifies the implementation of multi-concurrent proposers.
2. Allows for more flexible fee structures and user relationships for applications.
3. Enables safe pre-commitments without compromising network safety.
4. Potentially solves the challenges of duplicate transactions and fund exhaustion.

While not perfect, especially for general computation scenarios, this solution offers a practical balance for many blockchain applications, particularly those focused on specific use cases or industries.

Facts + Figures

• Commonware is developing a library of MIT Apache II primitives for product-driven blockchains
• Patrick O'Grady was previously the head of engineering at Apple Labs for 3.5 years
• O'Grady also worked on crypto engineering at Coinbase
• The presentation was inspired by a recent blog post by Anatoly Yakovenko and Max Levchin on multi-concurrent proposers
• Traditional blockchains use a sequential process of build, replicate, verify, and then build again
• Solana introduced streaming state machine replication to reduce latent safety in blocks
• DSMR allows all validators to act as leaders simultaneously, potentially increasing throughput
• The proposed system pre-sells network capacity to out-of-protocol sequencers
• Accounts are allocated to exactly one sequencer at a time in the new system
• The system requires M of N committee certificates from sequencer broadcasts for a block to be valid
• Commonware has developed 25 primitives for building product-driven blockchains
• An optimized version of Simplex, similar to Alpaglow, will be released by Commonware

Top quotes

"We're building a library of MIT Apache II primitives for product-driven blockchains."

"If you want to be a decentralized exchange or market, you need to run auctions. And if you prioritize cancels, you need a way for participants to be able to get transactions in that take advantage of that order prioritization."

"DSMR takes it a little bit further and says, instead of having one leader, what if everyone is a leader and you build, replicate, and then verify each other's stuff and then eventually find a way to stitch it together."

"The benefit for an application is if you are yourself a sequencer, which makes a lot of sense, you can determine ordering for your application, but also determine how to charge people for fees based on what they're doing."

"We think that this is a pretty good balance, and the really big benefit you get is that because these sequencers have this direct relationship with the users, they can also offer like pre-commitments in a really safe way without actually occupying any safety from the network."

Questions Answered

What is the main problem that multi-concurrent proposers aim to solve?

Multi-concurrent proposers aim to solve the issue of market makers being vulnerable to exploitation in decentralized exchanges and markets on blockchain networks. By allowing multiple validators to propose blocks or chunks simultaneously, it ensures that market makers can cancel their orders and get ahead of potential adverse trades, protecting them from having their positions exploited between batches.

What is Decoupled State Machine Replication (DSMR)?

Decoupled State Machine Replication (DSMR) is an approach to blockchain design where the processes of building, replicating, and verifying blocks are decoupled and can occur concurrently across all validators. This differs from traditional blockchain designs where these processes happen in a strict sequence. DSMR potentially offers higher throughput by leveraging parallel processing, although it introduces some additional latency in block settlement.

What are the main challenges in implementing multi-concurrent proposer systems?

The two primary challenges in implementing multi-concurrent proposer systems are duplicate transactions and fund exhaustion. Duplicate transactions occur when multiple validators include the same transactions in their proposed blocks, leading to wasted throughput. Fund exhaustion happens when an account's balance is depleted in the first transaction, rendering subsequent transactions from that account invalid due to insufficient fees. These issues can significantly reduce the effective throughput of the network and create opportunities for exploitation.

How does Commonware's Sequencer Driven Decoupled State Machine Replication work?

Commonware's Sequencer Driven Decoupled State Machine Replication pre-sells network capacity to a set of out-of-protocol sequencers for each epoch. These sequencers are responsible for allocating capacity to accounts under their management. Accounts are allocated to exactly one sequencer at a time, preventing duplicate transaction submissions. Sequencers maintain the fee-paying relationship with the network, simplifying fee management. This approach effectively outsources the complexity of managing multi-concurrent proposers to the sequencers while maintaining censorship resistance and fair ordering.

What are the benefits of Commonware's approach to multi-concurrent proposers?

Commonware's approach offers several benefits: it simplifies the implementation of multi-concurrent proposers, allows for more flexible fee structures and user relationships for applications, enables safe pre-commitments without compromising network safety, and potentially solves the challenges of duplicate transactions and fund exhaustion. Additionally, it gives applications acting as sequencers control over transaction ordering and fee structures for their users, allowing for more tailored and efficient blockchain interactions.