DAPPCON 2025: Marc Harvey Hill - Surge: Accelerating Ethereum's rollup roadmap

Transcript

Hello. Um, my name is Mark and I work as an Ethereum core developer on the Nethermine client. Um, today I'm going to be talking about Surge, which is a base rollup template template built by Nethermine. Um, and I'm going to be giving a high level overview of Surge and how it's accelerating the Ethereum uh, rollup road map. Um, and a big thanks to David Leonardi for making the original presentation this is based on. Uh so roll-ups are just fancy multi-IGs. They don't really inherit Ethereum security and roll-ups are just slow. Um so maybe this is a legitimate criticism of uh Ethereum's rollup landscape today.

Uh but really we see this more as a challenge. So how can we actually build uh a rollup that is more neutral uh trustless all while remaining highly performant and secure. Um and that's really the aim of Serge. Um so I'm just going to give a quick kind of uh background on how uh rollups worked um or sorry do work traditionally. Um so the high level motivation for roll-ups is there are limitations to how much we can scale Ethereum's base layer um all while keeping it kind of uh decentralized like the throughput can only go so far. Um so the idea behind rollups is essentially move all this work offchain. um we're going to process the transactions offchain and then

ultimately post the um the proofs and transaction data to the L1 um where that can all be verified by a smart contract. Um and traditionally or most of the rollups today uh work with a centralized sequencer. Um so this is just one central entity which is going to be responsible for um collecting all the the users transactions um ordering them and ultimately posting it to the L1. Uh so the problem with this approach is uh there's just going to be one uh single uh point of failure. So if the sequencer goes down um then you're not going to be able to get your uh transactions on chain. Um and this is also a censorship vector. So if this sens uh if this

centralized sequencer decides they want to ignore your transaction um then they're just able to censor you freely and this concentrates power uh with this centralized sequencer and it just relies on the reputation of the entity running the sequencer. Um, one kind of caveat to this is that roll-ups can have a forced inclusion mechanism. Uh, but generally this is going to be uh it could be more expensive. It will go through some kind of uh waiting period. So you can't get your transaction on chain immediately. Uh and you could also be exposed to some kind of toxic MEV extraction. Um so how can we actually do better than this? And this really is the idea of a base rollup. Uh so kind of the high level

idea here is get rid of the centralized sequencer and instead we're going to uh essentially use the Ethereum L1 proposer um as the sequencer uh and just reuse a lot of Ethereum's proposal builder architecture um instead of introducing new infrastructure. Uh so the way this would look is if you are um sending your L2 transaction, you're going to send that to a separate mempool. um L2 searchers are going to bundle these L2 transactions and then they'll be sent to the L2 block builder um who is actually the same entity as the L1 searches. So once the L1 searches then bundle these transactions, they'll go to the L1 block builder and then the block will be

proposed. So yeah, essentially here we're just reusing a lot of Ethereum's um existing proposal builder architecture. Um and this gives some uh significant benefits. So for one, we're going to inherit in uh Ethereum's livveness and censorship resistance properties. Um again, we're not introducing any additional infrastructure. So this just generally kind of simplifies everything and is a much more cost effective design. Um it just generally aligns everything more with Ethereum's uh base layer uh and can also give um some composability benefits between different base rollups and with the Ethereum L1 as well. Um, so with Serge, our aim is really to

build a neutral rollup. And what do I mean by that? So I'd say a neutral rollup is something that is permissionless trustless censorship resistant, open source um, and something that anyone is able to deploy and um, build on. Um, and all of this while still uh, being extremely high performance and secure. Um, and that is really our aim for surge. So surge is a base rollup template built by nethermind. Um it is stage two at launch. Uh meaning it has very strong livveness censorship resistance properties. It is based um it is uh uses ETH. It doesn't introduce any other token. Um it burns 100% of the base fee and it has giga gas performance. So can

uh process billions of gas per second in throughput. And in order to do this um we've deployed on the Tao Allesia stack which is the most mature uh based rollup stack at the moment but we've made uh several modifications. So we've replaced the finality gadget. Um so now we have a two of three uh multirover system which I'll speak about a bit later. Uh we've gotten rid of the bond token which was kind of tao bond. Now as I said everything's just with ETH. Uh we've changed the fee split. So uh at the moment everything is 100% burned although this is still subject to change on mainet deployment.

Um we've replaced the execution client from rollup g to nemine client and we've just expanded the documentation a lot just to make it a lot easier for anyone to kind of uh deploy their own surge instant and surge is stage two at launch. So what does this mean? So um stage 2 is defined on this website L2B which maybe you're familiar with which kind of classifies different rollups um according to their trustlessness and security guarantees that you're getting. Um and on Serge we have a 45day exit window. So that means if there's some kind of upgrade happening then you'll have 45 days in which you can exit your funds. Um and there are no kind of

emergency powers and there's very limited governance. So essentially like we're saying, you don't have to trust us. Um unlike a lot of roll-ups today where you might have some kind of emergency powers where they can just kind of make some changes with not a lot of notice. Um so it's just not really a trustless system. Um and yeah, so it is fully tokenless. So everything is based on ETH. There's no external token introduced which is going to introduce some kind of um outside economic incentives or speculation. Um, and it's just generally very aligned with the L1.

We don't have the kind of overheads of introducing some kind of new token and distributing it, etc. Um, and yeah, it has extremely high performance um, thanks to the Nethermine client um, which so yeah, we've achieved a gigas per second throughput um, as has been demonstrated on the base main chain. Um and this is all thanks to a lot of kind of uh crazy technical optimizations uh done by my colleagues also working on the nethermind client. We have a two of three uh threshold system for proving. So we prove with u a te based system um sp1 and risk zero and we need a quorum of two out of three of

these provers to agree uh for the proof to be valid. So this just generally um removes having one point of failure uh by just having kind of one proving system because now we have kind of a lot more security and fault tolerance uh with this two of three quorum. So yeah you can get started like look into surge today. So check out this website surge.wtf uh you can kind of experiment with deploying your own kind of part of the surge testn net. Uh you can check out the docs or contribute to Serge on the GitHub or you can just start playing around on the test net. You just need to um get some hoodie ETH from the faucet and then bridge it over to Serge and

then you can start trying that out today. Um and we have kind of several powerups and improvements or power-ups or improvements as we call them um in the pipeline. Uh I'd say the biggest one probably is pre-confirmations. Um so we're aiming to have uh uh to preconfirm transactions on the order of like a couple hundred milliseconds. Uh we are working on improving this uh prover security and the block building and also kind of really unlocking the benefits of the composability that both uh base rollups can provide um by sort of having like a unified liquidity layer synchronous composability between different base rollups and between uh the L1 also.

Um so what's next? Um, so Serge is a base rollup template. So really it's not designed for end users, but it's really focused on kind of making something for other builders to kind of play with. Uh, so we really want the community to take over Surge and kind of deploy their own initiatives uh based on it. Um, yeah, we're going to continue to improve the proverality gadget, make kind of different technical improvements and of course deploy surge on mainet. And one such project uh which is planning to be deployed uh on top of Serge which may maybe you've heard of is Ethereum R1. Um so this is a neutral rollup initiative. Um so it's fully uh funded by donations uh kind of run by

builders uh and just a a public good. So yeah so it's I think the plan at the moment is to raise around a thousand ETH. Um and then all of the fees that are generated are also going to go to the uh research and development um of the of the chain and then after 2030 all the fees are just going to be burned. Um so yeah it's just uh essentially a communityr run initiative uh community governance um and this is really the kind of ethos of search. So yeah, you can check out this uh website, scan this link to try out surge. Um and yeah, sort of on this topic of accelerating the road map, I wanted to also mention another project

we're working on at Nethermind uh which is maybe kind of like a broader thing. Um we have this project way which is about automating the Ethereum protocol development. So this is kind of uh on the layer one as well as like the whole ecosystem. So accelerating surge rollup or different rollups. Uh the aim here is to have uh autonomous kind of core developers that can take over parts of my role in terms of development also kind of act as a coordination layer uh between different parts of the Ethereum ecosystem and and rollup ecosystems so that in future you know like it'll be even easier to kind of uh deploy your own roll up your own kind of and build more neutral u rollups. Yeah. Um yeah uh yeah this is the talk and yeah

you can check out search here. So thank you very much for listening [Applause] questions from the guest person. Thank you. Um nice talk. Thank you. Um yeah, I had a question. So you have this three tier for like two out of three proving system. Yeah. Um and two of them run zero and SP1. Yeah. I assume this is not uh Nethermine running in there. Um no, not as I don't I don't think so. So it's it's okay. Okay. My question is

like I assume it's got to be rest because as far as I know that's the only thing that works in EKVMs at the moment. Uh but then how do you confirm the compatibility between the two? Like how do you say okay you know uh two of them are running rest one of them is running nethermind presumably. Uh how do you uh yeah how do you resolve if there's a a bugs you know like a consensus bug because there are two that are pretty much the same client and the last one is a third. So, Nethermind is always out uh you know um out voted basically.

Um well, I'm not too sure to be honest on the details of whether Neil sorry, Nethermind or WH are running in them, but as long as any two of them agree. I mean, they could disagree not necessarily based on the thing, but on the some kind of bug in the prover system itself, right? So, I think in those cases, it would still be helpful to have three. But I think this is one thing that we are working on is like improving the general resilience of the approved architecture. There was another question. No, so I assume search isn't perfect either.

What would you count as major downsides of it? Um, I mean at the moment I guess one of the bigger disadvantages of base rollups is not having pre-confirmations, but as I say that is in the pipeline. Um, I mean I can't think of too many other downsides. I think it's quite a big improvement. Um, do you have any numbers on proving costs? Uh, no. Okay. Uh so what happens if no two out of three provers can agree on on the

output of a specific block and uh and you can't come to consensus is all stuck on the rollup. Um I think it would be a big problem if two of three don't come to consensus. Uh yeah I think this is going to just be a fundamental issue. So you can like resolve it I guess by having even more redundancy or kind of but yeah I mean that's the whole thing of like stage two is having minimal intervention governance right but then that always comes at the cost of like in an extreme scenario if the provers can't agree then like you're kind of stuck right thank you for the talk you the

incentives to that that existing layer 2s like B not base maybe but optimism and arbitron maybe also base switch to a a base rollup model or is it like network effects that if everything is just layer one so to say or yeah yeah exactly yeah so this kind of idea of um you call it like synchronous composability so this allows you to unify liquidity between rollups so you can imagine in a future where there are many like base rollups then you're going to be at quite a big advant a disadvantage to be sort of outside of that unified ecosystem right you're going to become more like fragmented and and island um so at the moment it's it's

I guess the base rollups are quite small but we can imagine that um potentially many rollups or kind of even smaller roll-ups could kind of group together and be based and like you know have have this in uh composability between each other and then of course tap into the kind have liquidity of the L1 as well. Yeah, I hope so. Yeah, maybe question for me then. Um the base rollup you mentioned like no pre-confirmations at the moment for base rollups, but it's coming, right? You said like it's Yeah, so this is in the works. So I think um we're aiming I've heard for around like 200 milliseconds is the aim for uh pre uh base

preconfirmations. But yeah, what's what's the trick? Like what's um to be honest like I'm not like super involved in the the pre-confirmation side. So but yeah, it's obviously a different challenge to with like centralized sequencer which is much easier to give uh pre-confirmations.