Yuan Su
Yuan Su is an experienced software engineer and a co-founder of Metis, an Ethereum Layer 2 Rollup platform. He is also the CTO and co-founder of Nuvo Technology Ltd., a Metis EcoNode designed to simplify blockchain management.
Education
Su graduated from Acadia University with a Bachelor’s in Computer Science in 2005. Later, in 2017, he pursued and achieved his M.B.A. at the University of Toronto. [1]
Career
While studying at Acadia University, Su interned with IBM, a global technology company. After graduating, he stayed with the company for over 14 years as a software engineer and technical lead. In 2016, Su met Elena Sinelnikova and Kevin Liu at a collaboration conference held by CryptoChicks, a blockchain education platform for women. Together, they co-founded Metis. [1]
While working at Metis, Su worked with CaseWare International, a global software provider for auditors and accountants, as head of development for Hybrid Cloud until 2023. During this time, he co-founded Nuvo Technology Ltd, an EcoNode of Metis designed to simplify blockchain management for its users. [1][2]
Interviews
Metis and Layer 2
In June 2023, NewsBTC interviewed Su about Metis and its involvement in Layer 2 development and infrastructure. After a brief introduction, he discussed about what Metis originally started as: [2]
“Layer-2 wasn’t our first mission at Metis. We initially promoted on-chain collaboration of DAOs and focused a lot on stuff like voting, treasury management, etc. But we soon realized it was equally crucial to focus on business activities—operations, internal transactions, etc.—which existing DAO frameworks did not support. We saw the opportunity to build solutions that’ll enable on-chain business activities, replacing off-chain entities that serve centralized power. But we hit a roadblock. The infrastructure for fast, inexpensive, and reliable on-chain operations wasn’t ready yet. That’s when Metis got into Layer-2. And this also points to why L2s are critical for Web3 in general.”
He then explained why Layer 2 solutions are necessary for Web3: [2]
“Layer-1 platforms, like Ethereum, aren’t scalable enough to enable full-scale business activities on-chain. Trying to achieve the desired level of scalability on the Layer-1 isn’t advisable either. One, it’ll have severe security trade-offs; two, it’ll be very very complicated. And you’ll anyway face restrictive gas fees and overall transaction costs on the base layer. L2s are thus mission-critical for building the Web3 Economy. However, the fees on most existing L2s is still too high to bring Web3 to the mainstream. Plus they often have centralized sequencers, making them prone to manipulation or censorship.”
When asked about how Metis achieved their low fees, he responded: [2]
“Memolabs deserves a big shoutout for this. Their decentralized storage layer was a great help. But I’m proud that we innovated Optimistic Data Availability which, despite the criticism at the time, is critical for the industry-low costs on Metis. It’s roughly 15 cents per swap and 5 cents per ETH transfer. We optimistically assume that sequencers will make the data available. If everyone behaves honestly, this ensures the lowest availability cost because you don’t need to upload the entire transaction data to the L1 all the time, paying gas and other fees every time.”
“That’s like an on-demand rollup which submits a fraction of the entire transaction data as and when required. To make this work, we innovated a unique sequencer rotation mechanism where the sequencers in a pool—each with a finite number of allotted transactions—take turns to serve the role. We’ll fully implement this soon, using threshold signing to reach a mini, pre-rollup data availability consensus among sequencers as the first line of defence against attack vectors like Denial of Service (DoS), etc. Metis also has a mechanism to pause sequencers through tiered community voting, which makes the cost of tackling rogue sequencers both finite and feasible.”
Su also explained how decentralized autonomous companies (DACs) functioned in the Metis ecosystem: [2]
“Metis has a contribution and reputation-based governance model that enabled the transition from DAOs to DACs or Decentralized Autonomous Companies. We developed this idea to address the various issues with popular token-based governance models: whale voting, early adopter influence, vote sales, and abuse of power by delegates. These models are also time-consuming, while voters often lack the experience or knowledge for informed decision-making. DACs are very efficient in autonomously handling business activities like administrative tasks, payrolls, communications, etc. Their governance model is based on the participants’ actual contributions and their reputation within the community.”
He then went into some of the challenges Metis faced during development: [2]
“Applying MIPS was our first major challenge. L2s usually have a computation C that the L1 must verify before finalizing blocks. We had to transform this C into an equivalent polynomial for our zkSTARK-based approach. But this posed questions regarding processor specifications, number of registers, applicable instruction set, etc. Plus combining Optimistic and ZK Rollups is complicated as their validity-proving methods are apparently contradictory. Implementing Optimism’s Cannon framework helped us overcome these issues, besides providing a MIPS-based alternative to the complex process of generating zk-proofs using EVM opcodes.”
“The next big challenge is high data storage costs—both on-chain on L1s and off-chain on L3s. This isn’t exclusive to Metis. Every protocol running smart contracts on L2 and bridging contracts on L1 faces restrictive storage costs at some level. But for us, it was also about ensuring high TPS despite the Hybrid Rollup framework. So we adopted a recursive proof method called Fractal Scaling—zkSTARK proofs on L3 help verify whether the transaction batch on L2 correctly computed the L1 state roots. This reduces costs because a single proof represents thousands of transactions per batch.”