Equilibrium Infra Bulletin #33: Block-STM vs Sealevel, Fraud Proof Wars, Primitives for Network Privacy, and more...
Equilibrium Labs builds the state-of-the-art of decentralized infrastructure. We are a global team of ~30 people who tackle challenges around security, privacy, and scaling.
🔍 Block-STM vs. Sealevel: A Comparison of Parallel Execution Engines
⚡️ Topic Summary
Block-STM (Aptos) and Sealevel (Solana) are two approaches to parallelizing blockchain execution. Their main difference is how they manage concurrent state updates, i.e. situations where multiple processes use the same resources:
Block-STM assumes (optimistically) that transactions are not dependent on each other and only re-executes transactions when there is a conflict (e.g. tx B is dependent on tx A). This model is simpler for app developers.
Sealevel (pessimistically) assumes that transactions will conflict and relies on lock-based synchronization to prevent conflicts. This leads to more upfront work for developers who need to define dependencies and declare what portion of the state they’ll be reading from and/or writing to during execution.
To compare the performance of both protocols, the team forked the Aptos core for Block-STM and the Agave client codebase for Sealevel. Both tests were performed on the same hardware: Latitude’s m4-metal-large (384 GB of RAM, CPU with 24 physical and 48 logical cores). From the graphs below (1k and 10k blocksize), we can see that Sealevel performs significantly better than Block-STM for both high contention (2 accounts) and low contention workloads.
While these results seem relatively robust, there are some caveats which are expanded on in the full post. The blog also covers earlier protocols that contributed to Block-STM (TM, STM, Calvin, and BOHM) and goes much deeper into the technical details of both protocols.
🤔 Our Thoughts
Parallel execution enables more efficient utilization of the underlying hardware and taking advantage of the general trend of computer hardware evolution (see graph below). However, it’s not sufficient in itself to meaningfully increase throughput - parallel state access is equally (if not more) important.
Most newer protocols seem to adopt optimistic parallelization or Block-STM directly (Monad, Sei, Starknet, Movement Labs, etc), rather than going for the pessimistic approach. Sealevel is currently leveraged by SVM L2s, such as Eclipse.
💡 Research, Articles & Other Things of Interest
🤓 Mova: Nova folding without committing to error terms: Mova is a new folding scheme proposed by the Nethermind Research team that modifies Nova to improve performance and only requires 3 rounds of communication.
📚 Trustless Interoperability Between Rollups: Landscape, Constructions, and Challenges: Fresh research from 1kx explores different approaches to interoperability and their tradeoffs through two illustrative examples: Cross-rollup token transfer (send to self) and cross-rollup token swap.
📚 Nightshade 2.0 Launches on NEAR Mainnet: With ZK sharding and stateless validation, Near validators no longer need to track all shards (lower hardware requirements). The focus going forward will be increasing the shard count (from 6 to 10 this year) and implementing dynamic re-sharding (scale nr of shards according to demand).
📚 Fraud Proof Wars: L2Beat dives into challenges with fraud proofs and how the approaches of Arbitrum, Cartesi, and Optimism differ.
🎧 Primitives for Network Privacy - Will Scott: A talk exploring network privacy, i.e. what can be inferred from observing communication patterns between different parties.
🤌 Personal Recommendations From Our Team
📚 Reading: The Next Fifty Things that Made the Modern Economy - Tim Harford: Covering everything from bricks to blockchains, and Santa Claus to credit cards.
🎧 Listening: Kasey Chambers - Lose Yourself (Eminem Cover): The recording of this live performance was stitched together from footage that fans had filmed.
💡 Other: Calling All Hackers: A Phrack article by CTS (co-founder of Zellic) about lessons learned going from a hacker to running a company.