The Urge to Surge: Rollups and the Quest for Max TPS on Ethereum | Sandy Peng - Scroll

[Applause] you all right well let's get started let's get right um going perhaps just um to start we have um many different projects and teams we've got um some of the the major kind of um l2s um here on this panel do you want to First say your name and then um the the project that you work for just so we give the audience a little bit of background on who's who's up here today hi everyone I'm Stephen I am from offchain Labs we contribute to the arbitrum technology stack as well as prism which is the leading consensus client for ethereum but here today I think to talk about arbitrum hi I am uh Ellie uh co-founder CEO at starkware and and uh we support the Stark net layer 2 uh hi my name is run uh I work at mantle we are an OP stack uh based rollup that uses offchain data ability using ion layer da hi everyone I'm hen from scroll so I'm the co-founder uh inside of the scroll and inside the scroll I lead the engineering team and the scroll is a uh ZK Rob uh evm equivalent ZK Rob for ACM amazing I'm so excited to to chat with you guys thank you for being here let's get started by talking about 4844 about the denune upgrade that um is really designed for l2s it's one of the first you know in protocol changes where ethereum's really hammering home the vision that wants ethereum wants to scale through l2s and not on the L1 um one of the the main kind of um questions around um EIP 4844 is um exactly how much of a benefit is it going to be to l2s um there's a lot of back and forth with ethereum core developers about how many blobs to introduce they finally landed on three blobs per block with a maximum of six um but I'm curious to know from um you guys like how prepared are you for this upgrade um how important do you think how much of an impact do you think it really is going to have on um on layer 2 adoption um H Chen do you want to go first uh yeah sure so I like 444 is still like a very important upgrade uh that come into the it's very important for the layer two uh because like you can save on the data availability by just using the blobs that is like the can be expired after two we about two weeks um but I think like the uh people may like initially overx uh being overly optimistic about how much like reduction of the 444 data Bloss it can reduce I think like the from the current test net I see is like the it's about like the uh 3 to 4X reduction compared to the call data in the saving the cost which like people initially think like it's maybe 10x but I think like overall it's still quite important because as like the uh layer 2 the uh the majority cost of the layer two transaction is still coming from for the data availability so like 80 90% of the that is coming contributing from the U being like making the data available so like using the 444 data blocks can still reduce a lot of under uh layer to transaction cost and then we are like working on to supporting the 444 very soon gotcha anything to add um Ain uh no I I think that's that's pretty accurate like like H mentioned 4844 is is foundational in a sense that it paves the way for full DK sharding at which point the uh the onchain ethereum based datability will increase spite order magnitude but the actual increase right now is going to be fairly mild at least initially while they're bootstrapping the network um like like like aent mentioned uh 3 to six data blobs which equates to about a max about 750 uh kilobytes of data per block which would equate to around 100 to 300 extra TPS across all rollups based on some reasonable assumptions about the uh the the number of bytes per per transaction uh based on transaction complexity and in terms of uh how this actually affects something like like mantle which do offchain da I would argue it it it opens up optionality for sure but since in almost all cases the the cost of posting data to an offchain da solution like I going order magnitude lower than anything with uh onchain call database data ability or data blood ability it's unlikely would change our our uh methodologies that much based on it that's fair just to pick up on that so I mean sure you know data availability costs are going down on ethereum through EIP 4844 but um mantle is still going to be be using other um data availability Solutions because they are still cheaper than what's going to be available on ethereum is that correct yeah for sure it'll it'll be dramatically cheaper at least for the near term gotcha um Ellie I know star where was one of the first to be ready for this upgrade um can you talk a little bit about um I mean picking up on on that point of you know you are using ethereum for data availability and and being very very conscientious about making sure we're ready to utilize the full benefits of 4844 as soon as it goes live um but why not um also experiment and and see kind of the the perhaps more performant or less costly da layers out there like Celestia and and other l1s oh yeah we don't object to using them uh we will be using data availability layers as well um so yeah no objection there to that we'll also have a volition at some point um you know you want to really integrate things and maintain proper security and proper interface with all of the infrastructure and the thing that our developer ecosystem wanted first to have is um uh this uh you know the Blobs of EAP 4844 because they basically get ethereum security just turned on with no change um once we bring in things like um data availability or evoltion it means that they would need to change parts of the code like to opt into uh using those things and it would be S slightly more work for them and uh but it really depends on what kind of cost reduction we'll see with the 4844 if it's as significant as a promised then we'll be happy with that for a while that's a thing I I think a lot is up in the air about how how much the cost reduction is going to be especially if we see um more activity on l2s um and one of the things that I want to highlight about the analysis that has gone into um how EIP 4844 is going to impact l2s is from offchain Labs um I believe uh one of your your colleague Steven had written a blog post explaining that EIP 4844 um will likely be more economical for the larger rollups and for smaller rollups they'll have to rely on secondary blob markets to um to basically auction off different parts of the blob because they're not going to use the full 120 gate kilobytes can you talk a little bit more about like those secondary blob markets and um who those markets are going to be run by a little bit more about like the third order consequences of EIP 4844 on l2s yeah absolutely so yeah that's one of the nice things internally so as I mentioned we build arbitrum technology we also our team was very instrumental in building the layer one side of this the prism team uh and building um uh 4844 support directly into ethereum clients we sort of get to see both sides of that and it helps us with our understanding of the ecosystem and yeah I think the thing that you're alluding to is this idea of uh if you can fill up a blob you'll get a much better economics than if you have a partial blob and sending partial blobs will not be great so there's this hypothesize idea that people will somehow coordinate to to share uh and come together I don't know exactly how that will play out and what that will what that will look like um but what I can tell you is uh for the arbitrum public chains um they do will are expected to be just fine and be able to fill up those blobs so they'll be able to um benefit really particularly Arbitron one which is the one that sends it data to ethereum they'll benefit from uh 4844 I haven't done the exact analysis I know there's a lot of great people doing it my Approach is basically it's like it's coming in a few weeks we're going to we're going to see what it is and we'll uh we'll um you know but but arbitrum technology will be ready there's a dow proposal already live um making sure that the chains are ready to adopt this as soon as it's ready and so it will go live basically uh on day one and uh I think it's going to be very important one thing I think is going to come interesting from this is I think it's going to highlight a lot of the differences in uh L2 fees and layer 2 fees uh a lot of the uh today as uh you know as we mentioned there's uh the L1 fees are dwarfing everything else so no one's really paying attention to that little bit on top so some layer twos are charging this much some are charging this much on top and I think when you squash the bottom that you zoom in and that becomes important and there's a very important conversation on sustainability arbitrium is not the cheapest a layer 2 but it's sustainably priced I think that's going to open up a very interesting conversation I agree how much of the fee savings will be sent back to the end user of the L2 and to what extent um is it actually um you know desirable that that you you make uh transactions on l2's like like you know Subs cost um especially if that encourages more spam or makes it more um unsustainable for the rup to operate that is a very interesting question um any any thoughts on that before I I move on to one of the the one of the other questions that I had though from from the other panelists on the topic of like once the fees uh for data availability are significantly low to to what extent do you give back all those savings to to the end user versus you know for the sustainability of the road up yeah I see like for us is we initially when we design the scroll we have like the multi-dimensional fee structure we have like lay one fees and also the execution fee that we separate so once we using the 44 data boss then that automatically will translate that the layer one data fee will reduce a lot and then that will be like I think saving a lot of cost directly that user can benefit from that yeah so show hands here how many people have participated or um had you know team members of your team go to something called the roll calls that have started recently in the last couple months they're um spearheaded by the ethereum foundation to improve the standardization and and kind of interoperability between rollups show of hands who's gone to these rollup roll calls SL have sent members of their team to the roll calls okay first question to you then Ellie um because this is really about bad boys I mean part of it is really about also encouraging evm equivalents and so it doesn't come I know one of the reasons one of the unique things about um Stark Stark net is that it's not as um aligned aligned and this meeting is all about how do we ensure that you know the protocols and the changes that are happening on the L L1 align with the l2s and the l2s Align with the L1 tell us a little bit about um you know that decision to be not as aligned to ethereum and is that a conscious decision not to participate in the in the roll calls so uh we build a different um we build for the long term we build what we think is the best technology that we can provide um you know we invented Starks Cairo uh Stark net Stone by the way let me take this opportunity to tell you if you have time at 4:25 at the coral stage we're going to have a very interesting announcement about proving technology breakthroughs so please if you have time go attend it um anyways we um wanted to I mean the story of Stark net is that first of all we built you know we built the first L2 scaling solution which was the starx system and when we started at it everyone was saying first ethereum doesn't really have a scaling problem and two certain L uh you know validity uh technology is not going to be useful for that three definitely it's not going to be Starks now it ended up being that pretty much all of the leading l2s using validity rollups have converged onto Starks and um it is actually considered the endgame and we're very proud of it um Cairo and the starkit stack are the Next Generation smart contract language I think you'll see a lot of convergence on that as well we have this tendency of saying things that at first people think that uh you know were like crazy or don't make sense but over time people realize they actually do make sense so you know we've been termed unaligned I don't think we are unaligned we just build a bit different and we think of ourselves as trying to do the right thing with our technology all of this to say I still apologize for not joining the roll calls we'll try to make sure that someone attends them because we do very much support ethereum and we want standardization among l2s of course on the topic though of maintaining evm equivalents and and conforming to kind of like the standards set by the layer one which is for example like only using a certain um you know pre-compile address format that this was one of the topics of the Roll calls to reserve a set of precompile addresses just for l2s to use um is this a very um important initiative that um evm equivalents and and maintaining alignment with um L1 in the future I mean is that is that something that um say um Stephen that you for see kind of being like a top priority for arbitra yeah I think coordination like that particularly if you are building an evm chain if you're not then you don't need to worry about some of these things but you do want to have that coordination because you know if you use some pre-compiled address and it means something else in a different ecosystem and then ethereum introduces a new pre-compile now you've already it's just going to cause like chaos so that level of coordination I think is important um when it comes to the general like evm equivalence or not I I think we take um at offchain Labs a very middle position which is you know and that's what we call evm plus we're building something called arbitrum stylus which says evm equivalence on arbitrum isn't going anywhere if you want to build the solidity it's going to have the best support equivalent support as best as you can but we also don't view that as like the as the sailing we view that as the floor that's Baseline support which we think is important because so many developers and so much code today written in that language but then we say can we do more can we appeal to developers writing other languages and that's what stylist does so it adds support on top of the evm for developers that want to write in Rust and c and C++ and allows them to do that uh synchronously and also get a 10x uh benefit but also it's not only the developer it's also their code so if you're a gaming studio that's spent the last five years writing a gaming Ed at C++ you can now directly call that onchain so our approach is evm is very very important because it has so much adoption in our ecosystem today but if you zoom out they're estimated to be about few tens of thousands of solidity developers about 3 million rust developers about 12 million c c and C++ developers and we say can we cater to those people too and that to us seems like a big win of course we should keep evm equivalents and not do anything with that so this is all extra and additional but for those who want it can we give them a path directly onto the Arbitron chain so they can work together with solidity you can actually have a single app that's WR in solidity and rust uh for those that want to opt in and use that we think that's important there are some that think like this is like Sac religious like evm is like you cannot deviate from it and I respond to them and say ethereum itself tried to do eism which is very similar a few years ago not only that like I don't think anyone in the world today if they were starting from a including metallic would say let's create the evm from a fresh slate like if we could you know do the same way today everyone has changes you'd make that's the way software Works you're locked into some of your decisions based on Legacy requirements but like there's nothing like religious about the evm there's a very important reason to keep it and we think it's important for the adoption but we're also happy to expand it without hurting it and I think that's sort of the middle line approach we take but but that's also kind of like the beauty of rollups that I think a lot of core developers want to see they want to see the most experimental the most flexible type of evm they want to see code changes that they can't get pasted on the L1 because of how you know you can't really move fast and break things on L1 you have to you they want to push the more experimental things to l2s but I think there's also some push back from L2 teams of being viewed as like the testing grounds for just layer one changes I mean a a room what do you think about the narrative that you know l2s should be kind of like the test Nets of ethereum and then you know eventually the ones that the the projects or like the protocol changes that really work out then we like you know um do that on the on the L1 um I I think I can argue from from two sides there the first is that uh if you're creating an L2 and you're creating L2 that's fully evm equivalent you're you're going to have at this point I think a much harder time getting adoption for it than if you were to try and experiment just because um like this is going to be the year I think people going to launch a thousand different Roll-Ups And if every single rollup is just an even rollup what's the differentiator they get to you to like be the RO people use like maybe you could be the the Kazakhstan rollup or something and or organ nation states or or ethnicity or religion things like that but I think a much more interesting approach would be to kind of organize based on some kind of technical advantage or or change in AR or things like that and to that point like there is nothing that's going to stop like ethereum can't stop these rollups from from doing this like at best they can standardize and say okay well make sure you don't you know use like any random prefix for your pre- compiles otherwise any kind of crossr of client that we want to implement is going to have a nightmare trying to deal with all this different stuff um so I think that is a better approach these RS will experiment and then ethereum can eventually choose to do this to implement the things which the rup seems to succeed at um after kind of pick the winners um and I think we're seeing L standardization kind of occur not just with the roll call but also in terms of like the more structured uh uh rip the protal of which the first was to implement this this promile that used to be called uh EIP 7 7212 uh which adds a uh a decreased gas cost for a vctor verification to allow for um ERC 437 wall to have cheaper uh verification costs on the uh on the Chain itself I mean h Chen anything you would add to this because Scroll of all the rollups is not to not to use too awful comparisons but scroll is probably like the most ethereum align because it's trying to be like a type zero like you know like a ZK evm that is completely can take the entire evm and just be able to like replicate it fully in its exact State um and and you know this is a long-term vision for the ethereum protocol road map to um you know one day even have that enshrined in the protocol so I mean I feel like I mean what are your thoughts that that l2s are kind of the testing grounds for what ethereum will eventually want to be yeah I'm I'm personally guys think I support like more experiments on the layer too so that I can uh try out more features uh for and then in order to like making those experiment uh for example the new pre compiles be more useful is that like every all of the layer toes RS like which are evm compatible that adopt the same uh same standard and same uh same protocol so that uh so that the developer all of the developing tool toolings so that that can be all used across different layer Toes that that would be like uh benefit the developers and also reduce the confusions to the developers when like you you need to be very careful like turn down certain flags for when you develop on different chain that's like a nightmare for developers I would say but like when they come back to the uh seem like on on the layer ones I think like our argue for is that layer one to behold this gold St golden standard that being like like very stable not like the moving too fast on the changes because like otherwise uh if you're like they're making lots of the drastic changes to the evm itself like pushing all of the layer to that trying to be even comble to be catching up on a lot of things and then some things like example uh certain certain precompile it could be or some changes to the evm will not be very ziky friendly we're adding even more cost like for the zy even to catch up to implement those things and adding actual overheads for that so I would say like the so the what like a layer one should be like really thinking about the how to maybe like serve the all the layer two RS better like to uh reduce the cost like the 44 data bus and then like the uh and in the future for the food D Shing so and then can driving this row up Centric road map like then to keep uh growing the layer to space you know that would be nice but I do believe that ethereum core developers are really gunning for some pretty major changes to the evm not that it'll happen in the near future but like ethereum object format account abstraction um we definitely probably going to get uh BLS promiles in the next upgrade um and and it it it seems as though the confidence that um in l2s is waning a little bit as of late there was a a tweet a couple of weeks ago by vitalic that said he was three times less confident in this idea that you know we can just push all the complexity to l2s because l2s are not as proven as the L1 um so making L1 more simplified as a DA is not something that he's as confident about doing as he was Prior um does that does that shock you Ellie does it what does that kind of surprise you like that sentiment that that confidence in l2s to be able to inherit the complexity and the risks of what you know ethereum um might otherwise not be able to do um because l2s could fail because l2s could you know um could could have a major bug that causes a bunch of people to to lose money Etc like does that surprise you no in fact I think that um it's inevitable that some el2 will uh you know will experience catastrophic bugs um and failures and part of the reason is I mean we have to acknowledge that some of the l2s do not have their core security turned on so the thing I'm most worried about is that some el2 just do not have security turned on and it's heading towards a huge catastrophe if not on the major L2 then some of the uh you know rollups that are going to be based on it there will be a mount goau like event on some of these l2s that have no security turned on and do your own research on whether your L2 has a security and where where you entrust your funds now having said that the teams that are um serious about security and some of them sit here I I'm not worried about that I'm not worried about that look things could have bugs you know ethereum experience bugs um serious teams have a way to overcome bugs and uh if the technology you're using does not have security turned on you do not want to put funds on it do your own research let's talk a little bit about more about security and how we Define Security on on rollups I'm under I'm under the impression that many um you know rollups don't have working fraud proofs so really validity proofs or really have um DEC all all dependent on a centralized entity to to produce blocks on the l2s these are you know major kind of shortcomings of most layer twos um is does that keep you up at night um Stephen like the the the lack of security and the idea that a mount gox level event Could Happen um on ethereum because of the lack of of security well if if I were building a platform that had those limitations they would keep me up at night and that's why explicitly like an arbitrum security is always always first you know I agree with the things that L said and I know that we are agree on a lot lot of those points uh that security being uh really primary and the most important concern and that's why we've done things the way we've done things and made sure that security proofs have been alive in this ecosystem um from day one I do think you know to to your point about vitalic tweet like and to L's point every L2 l2s is not like one thing particularly now there's this explosion of layer twos and uh everyone has its own considerations and and it goes together those teams that take security seriously will also be very slow to do any upgrade that is like in a security conscience and won't be experimenting so you want to make sure you're in a secure environment today and one where you have serious developers that are making sure to to continue to secure that it gets much much harder in the world where uh now we're encouraging others to use this technology to build their own rollups and make their own decisions so you have to do your research not only on the core stack but also on those that are implementing it and I think um it's important though to make sure the core stack having security is the is the Baseline important thing and then then you also have to understand who's operating it and you know are you ultimately secure there so um doesn't keep me up personally for the users of arbitrum but as an industry I do agree that it's it's uh and one that cares deeply about ethereum I think it is something that that is very very scary and I'm personally concerned about and hope that uh we can solve soon gotcha gotcha maybe as a a final question because we now have to wrap up our panel but for all our panelists what is kind of like the one um major objective or or Milestone that you'd like to see your rollup kind of achieve this year let's start with scroll first um I think like the oh um for the scroll um so one mstone is like I think the next will be uh bringing like the using the 444 to continue reduce the uh reduce the cost on on the scroll uh and then the second thing like we are working on like to bring the more security features as I think Stephen said like we take the security first approaches so we're trying to build like a a multi-prover approach that we are bringing in like the uh one more like sjx Pro that's going to test alongside with the ZK proof and then after that we get working on like towards the decentralization I think like be important things like for the layer too I think we're starting from the approver to be having a decentralized and then moving up to the sequencer that that's a lot to look forward to this year Mel go ahead probably not only this year but I guess yeah I think building a rollup is hard there's many different objectives so I could speak myself personally I would love to see more standardization round fourth choice rule when dealing with offchain da uh mainly how does a how does the rup decide whether data is May been available to the L1 and what state would you consider e canonical and kind of like how this actually resolved at the L1 level because there's a lot of ambiguity right now I'll say uh performance and costs are going to be dramatically improved in Stark net I would love to say more about what we're expecting this year but I'm not going to spill the beans do go 425 to Coral stage and you'll hear what I mean um I would say two things are one developer experience which ultimately leads to better user experience so arbitr stylist I mentioned I'm hopeful that hits mayet of many Arbitron chains this year um and I'm pretty you know pretty optimistic about that and the second one is also continuing on the security and decentralization um points so uh vitalic has this uh sort of three-point rubric 012 where arbitrium is uh a leading um general purpose platform today as a layer one as a as a one it doesn't sound too exciting you say we're number one but like most are in zero so it's like pretty exciting but I'd like to see that go from one to two and also see things like arbit and bold deploy to main that which opens up validation for everyone uh security again is is critical and I I'm happy with where we are but there's still ways to go thank you guys thank you so much to the panelists and thank you everyone for listening thank youever okay sorry thank you so much thank you thank you so