Mysten Labs & Sui: vs. Aptos, other L1s

Circa 2 weeks ago, a lot of excitement was generated on the back of the media ? publishing content on Mysten Labs’ upcoming Series B fundraise. This led to a slew of threads on Mysten Labs’ Sui native L1 blockchain. This occurred in conjunction to me publishing my piece on Aptos . On surface, the two have striking similarities — most notably because both were incepted by teams comprised of ex-Meta executives (fka Facebook). As such, I thought it was apt to do a comparison between the two.

The caveat here is that it’s unlikely possible for a like-for-like comparison until both L1 blockchains have reached mainnet. Each L1 has disclosed different pieces of information, e.g. Aptos tokenomics will not be known until its mainnet; while Sui has released their tokenomics whitepaper.

Here goes…

Mysten Labs & Sui overview

Mysten Labs was formed by the Novi Research team at Meta — also worked on the Diem blockchain. Like Aptos, Sui is also not associated with Meta whatsoever, despite being formed by former Novi-Meta personnel.

Mysten Labs has designed its own high performance Layer 1 (“L1”) Proof-of-Stake (“PoS”) blockchain, Sui. While Sui attempts to solve the blockchain trilemma, it is centered around scaling composable and dynamic NFTs for broad metaverse applications including gaming, social, commerce. By comparison, Diem was originally designed to handle light payments traffic between a small number of custodial wallets (10s to 100s). Diem’s original architecture would not have been able to support mass adoption of any meaningful measure.

In addition, the group has also formed partnerships with protocols to apply its research on scalability and consensus improvements. The partnerships entail open sourcing and deploying its Narwhal & Tusk consensus mechanism (discussed below) as well as Move’s programming efficiency. Thus far there are two formal partnerships — Celo (a mobile-first, EVM-compatible network) and Sommelier (a transaction co-processor for Ethereum).

Team: Prior to incepting Mysten Labs / Sui, the co-founders had all worked together at Novi:

• Evan Cheng, CEO: Formerly Head of R&D at Novi and Director of Engineering (Programming, Languages & Runtime) at Meta; previously Apple for 10 years
• Sam Blackshear, CTO: Formerly Principal Engineer at Novi, focusing on the Move programming language
• Adeniyi Abiodun, CPO: Formerly Product Lead at Novi, Meta; previously VMware, Oracle, PeerNova, HSBC, JP Morgan
• George Danezis, Chief Scientist : Formerly Research Scientist at Novi, Meta; previously Chainspace, Microsoft
• Kostas (Konstantinos) Chalkias: Formerly Lead Cryptographer at Novi; previously R3, Erybo, Safemarket, NewCrypt

Anecdotally, it appears the Mysten Labs team has managed to recruit more of Meta’s crypto human capital. Team continuity and maturity is always an appealing factor to any start-up, which Mysten Labs benefits from.

Fundraising history

Series A

• Date: December 2021
• Fundraise, round: $36m, series A
• Asset type: Equity
• Notable investors: Led by a16z alongside NFX, Scribble Ventures, Redpoint, Lightspeed, Electric Capital, Samsung NEXT, Slow Ventures, Standard Crypto, Coinbase Ventures, others

Series B

• Date: Currently pending; leaked into media in July 2022
• Fundraise, round: $200m, series B
• Asset type: Equity
• Notable investors (rumored / known): FTX Ventures

Proforma cumulative funds raised of $236m.

Tokenomics

Whitepaper summary:

• Ticker: $SUI
• Supply: 10b fixed; no token emissions. A share of $SUI’s total supply will be liquid at mainnet launch, with the remaining tokens vesting over the coming years or distributed as future stake reward subsidies.
• No deflationary / burn mechanism
• Utility of token: 1) staking for network security, 2) transaction fees, 3) storage fees, 4) governance voting, 5) medium of exchange as Sui’s native asset. More below on 2) and 3).
• Further details to be announced (“TBA”)

Similarities to Aptos

Apart from both teams spinning out from Meta, there are some superficial and fundamental similarities between Sui and Aptos, summarized as follows:

1. Vision to solve the blockchain trilemma
2. Premised on Move as its native programming language, though Sui has a slightly different object model to Aptos
3. Parallel transaction execution and consensus (enables low latency and higher throughput) based on Block Software Transactional Memory (“STM”) — instead of ordered transactions and sequential execution of blocks in other L1 designs
4. State synchronization for optimized data availability
5. SDKs, APIs for improved DX / UX
6. Cap table of tier 1 venture investors
7. Valuation (more or less)

The aforementioned overlapping features have been discussed previously in my thread on Aptos, including benefits of Move.

Merits (& deviations to Aptos)

Based on the publicly available information disclosed today, the key differences upon drilling down further lie under the hood. They appear to be as follows:

Gas mechanism: Sui is able to keep gas prices low and predictable while incentivizing validators to optimize transaction processing and preventing denial-of-service (“DoS”) attacks.

At the start of every epoch, validators vote on a network-wide gas price used as a reference for users to anchor when submitting transactions for processing. Validators are provided incentives to keep gas close to the reference price per epoch, as well as being responsive. Those who submit high prices are penalized with discounted rewards. Likewise, unresponsive validators receive relatively less rewards and risk reduced delegated stakes, since users want transactions processed efficiently. Further, predictable gas leads to predictable transactions being processed. Gas prices are able to remain low even during periods of high activity as its throughput is able to scale linearly with more machines per validator; validators can commensurately add more machines to meet increased network demand. Traditional blockchains are often designed to run on a single machine per validator (or even on a single CPU).

A unique feature in Sui’s design is that storage fees are paid separately to transaction execution. In principle, storage costs are meant to be off-chain dollar denominated for storing arbitrary on-chain data. It is priced exogenously, determined through governance polls. From the user’s perspective, they don’t distinguish in fees paid. However, when user submits on-chain data, they pay for both gas and storage. Storage portion of fees are deposited into a “storage fund” which covers costs for validators to store users’ data. The fund is used to subsidize future increases in storage costs as the network grows and matures. Conversely, users can delete on-chain data in lieu of storage fee rebates, and they do so up to an economic optimum.

Below is an overview of Sui’s tokenomics:

Transaction execution: Broken down into two paths to consensus for the following transaction types:

• Simple transactions: Defined as independent with no causal relationships and employs Byzantine Consistent Broadcast on transactions. It is a leaderless protocol, eliminating global consensus’s overhead without sacrificing safety & liveness guarantees. Faulty validators do not impact performance in any significant way, whereas most leader-based blockchains which experience a crash in one or multiple validators see throughput fall and latency increase. Transactions are optimistically validated and executed individually in parallel instead of sequentially or in batches, as with most traditional blockchains. This is done almost immediately. Majority of transactions are of this nature, e.g. often users simply want to make an asset transfer to a recipient, where the only data required is the sender’s account and no interdependencies with other arbitrary parts of the blockchain state.
• Complex transactions: Defined as interdependent / intertwined with shared objects and employs variants of the Byzantine Fault Tolerance (“BFT”) protocol. Sui orders them and executes sequentially by using Narwhal (more below). Transactions are validated in 2–3 seconds.
• Transaction dependencies need to be explicitly stated. The consensus sub-system also scales in the sense that it can sequence more transactions by adding more machines per validator. This allows Sui to scale.

BFT consensus: Known as Narwhal & Tusk. Similar to Aptos, Sui’s consensus protocol is also a derivative of HotStuff. Both Sui’s and Aptos’ consensus mechanisms minimize communication required between validators to process transactions to achieve lower latencies. AptosBFT is partially asynchronous. Sui’s design to achieving consensus distinguishes the protocol’s mempool from the consensus layer, whereas most PoS L1s have monolithic consensus protocols.

• Narwhal (mempool protocol): Mempools store unconfirmed blockchain transactions until they are validated by the consensus protocol. Narwhal’s task is to broadcast transactions to consensus while maintaining data availability. Narwhal is independent from the consensus protocol and is able to tolerate asynchrony in forms of faults or intermittent loss of liveness.
• Tusk (asynchronous consensus protocol): Responsible for the ordering of transactions. Tusk uses shared randomness to eliminate additional communication between nodes and allows each node to determine the total order of transactions. When composed with Narwhal ensures high performance when faults occur.

Narwhal & Tusk can be composed with other Byzantine Fault Tolerance (“BFT”) consensus protocols. The team open sourced this code to allow the broader crypto community to benefit from their research (starting with their Celo and Sommelier partnerships).

In theory, the design has no known upper bound to its throughput scalability. Since there are no testnet results, it will be interesting to see how this performs in practice.

Sui’s security properties are able to withstand bad network conditions, network partitions or DoS attacks on validators because it doesn’t impose any synchrony assumptions on the network. In contrast, sustained network attacks on synchronous blockchains (i.e., most proof-of-work based blockchains) can lead to double-spend of resources and deadlocks.

Hardware requirements: Recall requirements to operate a full node can provide a sense to the degree of decentralization, though it’s too early to determine at this stage. Below summarizes requirements for Sui, Aptos, other L1 peers:

As of June 30 2022, >5,000 Sui nodes were running in 65 countries. Other factors which drive decentralization will be discussed at a later stage when both Sui and Aptos have launched mainnet.

Sui Development Kit (“SDK”): Though Aptos also has SDKs to improve DX / UX, Sui’s SDK makes interesting attempts to bridge other ecosystems and non-crypto use cases.

• Gaming APIs: Connects gaming developers directly to the general purpose L1, instead of a sidechain or gaming-focused L2. Gaming developers would then be able to interact with Sui’s ecosystem seamlessly (users, other dApps, assets) and the Move language.
• “SuiEcho”: Allows dApp developers to bootstrap communities by facilitating portability of digital assets from other ecosystems into Sui’s. E.g., Ethereum based Bored Ape Yacht Club NFT holders can mint Sui-based equivalents via SuiEcho, using their ETH-native PFPs as proof-of-ownership. While the Sui minted NFTs are independent assets, their avatars could then be used in Sui’s network. Likewise, it allows Sui dApps to tap into communities in other networks. While high-value NFTs can be bridged elsewhere, the value ultimately accrues to the original chain. Whereas, this approach allows Sui and newer L1s to benefit from existing communities captive in another L1. Aptos SDK has NFT features too but is currently limited to those native to Aptos.
• “Handshake”: Front-end tool that serves as a channel for users to distribute, claim / redemption of Sui digital assets (e.g. payments, merchant coupons) to both crypto and non-crypto users. For non-crypto users, no upfront account creation nor interaction with the underlying blockchain is required. Handshake’s roadmap includes onboarding of non-crypto users.

Performance: Sui has only commenced its incentivized testnet since June 29. I will update this benchmarking again once more testnet data is available. Otherwise, the following reference points for throughput available include:

• 120k TPS achieved from an unoptimized 8 core Macbook Pro
• 130k TPS when Narwhal was composed with Hotstuff, cited as empirical evidence in Mysten Labs’ research on consensus
• 160k TPS when Narwhal was composed with Tusk

Aptos’ Avery Ching stated, however, these throughput numbers do not represent end-to-end blockchain throughput as they only take into account consensus, network, and partial storage considerations. They do not factor in other important considerations such as transaction execution times, account access patterns, or authenticated data structures (e.g. Merkle trees) in a production blockchain. On the finality side, block times often get mistaken for finality when in fact, block times are an input factor to finality.

Ecosystem: Refer below for a comparison across Github stargazers, Discord members and Twitter followers for key alt L1s, as of July 25. Note Ethereum has not been included because its dominance is widely known; also because these new L1s have been informally deemed as “SOL killers”.

Sui dApps:

• Sui Wallet (native wallet)
• Ethos (wallet)

Aptos dApps:

• Other Aptos grant program participants: Martian DAO, Solrise Finance, Protagonist

Final thoughts

It’s still very early days for both L1s and the smart contract market is undoubtedly fiercely competitive. Both companies were incepted around similar times; similarly the L1 projects themselves were several months apart. Despite there being a lot of overlaps between Aptos and Sui, there are some interesting technological differences as well.

Technology is fascinating but it needs to ultimately solve meaningful problems for people to justify paying for it; technological innovation needs to be a sufficiently significant improvement for people to undergo switching costs from existing technology. It is also not the only single determinant factor of success — Jason Choi stated, technology superiority alone almost never wins — success also depends on business development, go-to-market. Technology needs to translate into performance then results.

When considering multiple facets of each ecosystem, Aptos stands out when compared to Sui as well as other alt L1s. In the short amount of time since launching, and despite having a few months of early mover advantage over Sui, its ecosystem is larger by:

• ~2x for community / users (proxied by Discord members and Twitter followers);
• ~3x for developer activity (proxied by Github stargazers);
• Similar throughput but mostly inconclusive;
• More metrics can be analyzed post mainnet, such as TVL, active users (wallet addresses), active nodes, minimum staking requirements, etc.

On balance, when layering in valuation at the time of writing, based on the latest rounds, Aptos appears to exhibit relative value given the ecosystem progress. Note Aptos just raised their series A round totalling $150m (cumulative funds raised of $350m) at a valuation of at least $2bn. For $2bn valuation, Aptos offers are more in terms of a robust community, developer activity, dApp ecosystem while being more advanced in its testnet. Mysten Labs / Sui would need to close the gap between itself and Aptos across every measure in order to justify its valuation. This would boil down to first principles of attracting quality developers.

Having said that, the team is credible and I have no doubt they will be able to execute their vision. They are well equipped from a pedigree, skillset and resource perspectives. While this thread has been more of a comparison between Sui and Aptos, Sui’s enhancements definitely fill a vacuum in the current L1 smart contract market. Should they successfully execute their plans, they are likely to be competitive following mainnet.

Several industry participants have desponded new L1 entrants, such as Aptos and Sui. However, the reality is that there does not exist a chain which is capable of mass adoption today. While the incumbents chains (L1s and L2s) certainly benefit from being early movers, the gap narrows as new entrants (Aptos, Sui) close in the gap — assuming they are capable of resolving substantial scalability pain points and providing strong DX / UX. Nonetheless, the market is likely to be large enough to support a multichain future. The key questions are what will that look like, and how many?

Sui’s testnet results should provide a better sense of performance, and relative to other alt L1s. Likewise, it will be interesting to revisit this comparison again post mainnet in order to analyze L1 performance / attritbutes across more metrics, including both of their tokenomics, vesting schedules, allocation across stakeholders. Time will tell how the underlying technology will drive overall DX / UX and who will attract more users / liquidity.