Growing Ethereum Through L2 Innovation | Ed Felten - Offchain Labs

good morning everybody I'm Ed from offchain labs and I'm here to talk about what we can do to at Layer Two to make ethereum better and grow the ethereum space but let me start by talking about why Layer Two why is it that we're focused on Layer Two as the place to innovate and grow ethereum uh this really goes back to uh vitalik's um post uh quite quite long ago on the rollup Centric road map this is the idea that ethereum would seek to grow especially seek to grow its scale not by trying to build a monolithic protocol that does everything but instead by focusing on allowing Layer Two and eventually later layer three and above protocols to uh to uh to grow by Innovation and to grow by horizontal scaling uh and this in to my mind is one of the smartest things that vitalic ever did uh which is really saying something um I think it's a brilliant move and it's one of the things that has set e ethereum apart so far and has laid the groundwork for future success why is this such a good idea well one of the reasons it's a good idea is that it follows the approach that was taken by the internet and the internet of course is by far the most success uccessful open credibly neutral protocol that has ever existed uh layering in the internet was one of the keys to success the internet swept the field and destroyed all of the monolithic protocols that were proposed as Alternatives even the ones that came before it and the reason for that is that by being layered uh the the internet could have each layer focus on doing one thing extremely well and it could allow Innovation at the higher levels of the stack because the layers pluged together you could pick a best-in-class solution at each layer and that could emerge naturally from the activity of people in the market and people innovating rather than trying to uh build build out everything in a in a centrally controlled way that's been hugely successful for the internet and I'd argue it's been successful for ethereum and has laid the groundwork for future success so how's it going layer 2 and layer 3 well if you look at the throughput of the layer 2 and layer three ecosystem in ethereum measured versus in units of ethereums what you see is that layer 2 and three collectively have about 30 times the transactions per second of ethereum about 80 times the gas usage about 40 times the call data and we have now more than 100 L2 and L3 chains this is huge success uh ethereum could never have gotten to the size eth Denver would not be as large as it is without the growth of L2 and L3 chains to grow the ecosystem and just to emphasize how much growth has come because of the L2 and L3 ecosystem here's a bar graph that shows the gas usage at different layers of the ethereum stack what you could see is that L1 accounts for only a couple percent now so in other words almost everything that happens in the ethereum stack today happens at layer 2 and layer three now that's not to say layer one is not important layer one is the Hub for liquidity and it's this foundation on which the security of layer 2 and Layer Three is built just like how there could be no internet without tcpip there could be no ethereum stack without layer one so it is the foundation it's critically important but it's not where most of the activity happens anymore that's why what if we want to grow ethereum we need to focus on growing at Layer Two and layer three because doubling or tripling or even 10 Xing the throughput of layer one is not going to make a big difference okay so how do we grow ethereum I hope I've convinced you that when we do that we want to be focusing on layers two and three but I want to talk more generally about what we need to grow ethereum what we should be focusing on versus what's just a distraction so let me tell you first a bunch of things I think we don't need we don't need silos we don't need different L2 and L3 Stacks that refuse to talk to each other we don't need exclusive clubs that you have to apply to join that you need to become a member of and sign away some of your rights to participate in I don't think that's the way forward either and we don't need rigged games we don't need short-term activities which uh which just exist to enrich a few people at the expense of most of the users of the system and you know what I'm talking about here what we need is to focus on sustainable growth growth that's open and welcoming and that brings more people in and connects more things together together rather than trying to put up fences so in other words what we do need is ethereum alignment now there's been a lot of chatter about ethereum alignment a lot of people accusing each other online of being not aligned I don't want to play that game I don't think it's productive what I do want to do is say what to me ethereum alignment means to me alignment means making the ethereum stack faster cheaper more reliable and easier to use for the benefit of using users and developers uh if we're doing that then we're ethereum aligned um and one consequence of this is that we don't have to accuse each other of not being aligned we can instead just each focus on being as aligned as we can because every developer every stack can be pursuing these same goals at the same time so I want to talk I want to structure the rest of my time around these themes of faster cheaper more reliable and easier to use uh and again emphasizing that this is for the benefit of users and developers we're not building a chain for the benefit of infrastructure providers I love infrastructure providers but the chain is not for them it's for the users and developers okay let's start with faster what do we need to do to make the ethereum stack faster there's a bunch of things we can do and should be doing uh and let me just rattle off a few of them uh first first we want to have benchmarks we want to have standardized ways to measure performance and realistic performance not just what is the largest TPS number you can say without bursting out laughing but what is a realistic workload what is a realistic set of transactions on a realistic State size so that we can actually compare and measure and know how well we're doing second we want faster execution we want faster virtual machines we want to execute the evm faster we want to introduce new models of execution that are faster but also compatible and uh and composable with evm uh we want parallelism we want execution engines that can discover parallelism and exploit it when it's available but without making developers lives much more complicated by making them think about parallelism in race conditions and all of that we want resource pricing that's better so that we can better align incentives of users and developers with the real constraints on noex execution and then finally we want to work on sharding to get horizontal scaling with less friction we need to do all of these things I want to talk about one particular thing that arbitrum is doing or a couple particular things that we're doing in the arbitrum space to make this so um the first of them is arbitrum stylus that is um really focuses on what developers need in a faster ecosystem you want familiar standard battle tested languages and tools you want fast execution and most importantly you want to be able to write in those languages and execute quickly while having full composability with evm that's what arbitrum stylus provides you there's uh examples here of a few teams who are building with arbitrum stylus basically this is a uh a feature that lets you write a smart contract in Rust C or C++ compile it and deploy it directly onto an arbitrum chain in a way that's fully composable with evm execution not executing in a separate Silo not needing a separate user wallet everything just works naturally so that users and developers don't even know whether the contracts they're interacting with are using evm or or rust C or C C++ the only way they know is that it's much cheaper and faster if you're using stylus I also want to talk about horizontal scaling what that means essentially is instead of having a single L2 or L3 chain you can have many chains at the same time time and there's been an incredible explosion of of different chains um here's a uh here's a diagram that shows some of the chains that currently use the arbitrum technology there are more using arbitrum of course there are others using other Stacks there are now well over a 100 L2 and L3 chains in the ethereum ecosystem and these not only give you growth in the execution capacity through and throughput bring down cost but also These Chains can be customized to meet the needs of particular applications use cases or user communities and that customization is also important that's key to that approach of of layering one huge advantage of layering is that different people can do different things at Layer Two or layer three uh without needing to lock everyone into a single model or a single configuration if you want to optimize for fast block times you can do that if you want a different economic model different governance model you can have that all while still being part of the ethereum stack in the ethereum community all right let's talk about cheaper how what's been happening at L2 and L3 to make execution cheaper for users but it's important when we talk about cheaper to distinguish two Notions of cheaper one is are we lowering the cost of actually operating the chain and the other one is just are we lowering the fees because you can lower fees in the short run well execution is expensive but that's not going to be sustainable to get a sustainable reduction in fees you have to start with a reduction in the cost of operation because at the end of the day VC uh sub uh subsidies are only going to last so long um and you're going to be in a situation where long run fees need to cover costs okay let me give you a great example of how the layered approach has led to a big cost savings um in operating the ethereum stack that's really passed through for users and that is the uh onset of data blobs in EIP 4844 so uh here is a very boring um page from uh ether scan showing a transaction in which uh in which the arbitrum one chain posted a bunch of its data but the important part is this which I've zoomed in on um that for this at for this particular batch of data from arbitrum uh it was posted at about a 40x cost reduction um on ethereum of course a cost reduction that passed through to users so how did that happen well this started by thinking about how can we lower the cost for ethereum validators and nodes of posting all of this data for uh L2 and L3 chains and people realized that l2s were posting call data on ethereum but ethereum call data did more than the l2s needed and so this new idea of blob data which just just did just what the l2s needed and was much cheaper for for ethereum nodes to provide so lower cost for ethereum validators that led to lower data fees on ethereum if you're using data blobs those lower data fees on ethereum led to a lower cost of l2s for posting their data and that in turn led to lower fees for L2 users a dramatic reduction in not only in fees but also in underlying cost by thinking carefully about what does lay two need from layer one and how can Layer Two provide what layer one provide what layer two needs at lower cost a big win from the layered design approach and from thinking about cost reduction in addition to things like this we also need to keep thinking about better pricing for the resources that we use fundamentally uh all of these chains use use gas prices or base fees as a way of slowing down the chain if the mechanism is worried that nodes are going to get overwhelmed so if we get smarter about data pricing then the pricer can hit the brakes uh less uh less aggressively because it can have more confidence that you're you're not overrunning the nodes so better pricing means more throughput on the same nodes also important not only for lowering cost but for increasing capacity okay let's talk about how to make the system more reliable and here I don't have any magic breakthrough research what I have is a lot of Engineers doing a lot of work uh and people thinking really hard and having the mindset of reliability and highquality software there's no magic bullet here this is people working every day to make the stack more reliable and I'm proud of what arbit uh arbitrum has done and what ethereum has done in that area ethereum is sometimes accused of being too careful about reliability and safety um and uh I think etherium uh should wear that accusation proudly because ethereum has been reliable it has been safe uh and it has continued to function well for its purposes um partly because the community works and thinks and tests so hard to uh to make it so so reliability is hard work by everybody all right now let me talk about easier to use because one of the things that one of the side effects that we've had as we've gone to this layered approach layer twos and layer 3es horizontal scaling with hundreds of chains one of the implications of that has been that there's some friction in doing things across chains or moving assets across chains of course A lot of people are working on this problem but right now a big part of the easierto use problem is not just people thinking harder about user experience but also working particularly on how to connect everything okay so here again I think we can take a page from the internet the internet of course has been hugely successful at getting all kinds of servers and services at different layers of the protocol all kinds of software products to work together in a really seamless way so what can we learn what does interoperation internet style look like and how can we learn from that in trying to get ethereum to interoperate well a internet style interop has a bunch of features that I think we could do well to try to copy one of them is it focuses on speed second is that it's standards based their standard ways of talking about what what is an address who are we talking to how do we encode the payload at each layer of the protocol each layer has its own way of encoding and representing data but it leaves to higher layers a lot of other decisions that belong there uh internet style interrup is asynchronous uh we don't have one node one service stopping and waiting for another instead we fire offer request to somewhere when the response comes back we uh we deal with it when it comes back asynchrony is really important uh in order to get good performance um and it's also really important for reliability because of the next couple bullets first it's uh next it's fault tolerant that is you don't just assume that the service that the chain that the node on the other end is going to be there you don't just assume it's going to be uh honest and you don't don't just assume that it's going to give you an answer that makes any sense at all all um every component of this system is prepared for other components failing going wonky going weird being super slow and finally it's security conscious you don't assume the party on the other end is secure if you're interacting with some other service some other chain you don't assume it's necessarily secure you can a user can decide to interact with a particular service or rely on it but you don't want one service to assume that some other service is secure without thinking really hard about it and without communicating clearly to the users what assumptions you're making and I think we'd be well advised in the as we build out interop in the ethereum space to try to emulate all of these features okay so what does that mean there are different approaches going on there's a really active discussion about standardization the EF ran a really productive interop offsite on Monday uh across town here in Denver uh but I want to talk in particular about we're what we're doing in the uh in the arbitrum community about uh interrup and you may have seen a a post on medium recently and a bunch of discussion about the universal intent engine this is the thing that we're really excited about building out I want to take a minute about what it means first of all this is an intent-based approach to interop so fundamentally what that means are actually let me just walk you through a sort of timeline of what the experience what the journey looks like like for a user um in this system so the user journey in this system looks something like this you have a user over on the left the user wants to do something so you might imagine the user has some assets over on chain a they'd like to move them over onto chain B maybe also exchange them to some other asset but at least move them so the user then um uh takes an action on the Chain where they have the asset which provides those assets into some escrow contract and provides some parameters I want move these to this other place here's how much fee I'm willing to pay here's how quickly I'd like it to happen Etc um all of that is the user's intent this is the this is what they're requesting from the system uh the use the application or the user's wallet then formats that and puts it into the universal intent feed that feed is publishes the fact that this user wants to do a particular move of assets from one place to another the user's funds go in escrow um eventually some party shows up some party who subscribed to that intent feed shows up says yes I'm willing to satisfy that over on the other chain I'm willing to give the user funds on the other chain here's the fee I'll take assuming that that matches up and I'll show the uh the the that provider's intent next then there's a cross-chain settlement protocol that executes the user gets their funds on the destination chain and their funds are paid out of escrow on the home chain the beauty of this is that it's flexible bullet can be very fast and trustless and the user experience can be really nice the user says what they want and then the rest of the mechanism is handled by their wallet or other service provider now the solvers Journey the party who's providing the other side of this is pretty similar they subscribe to the intent feed they decide they want to bid on this intent if they're chosen they put funds into escrow they participate in the cross chain settlement protocol their funds will be paid from escrow out to the user and then receive funds on the other chain later thanks to cross chain settlement so that's interrup by doing these things together we can align the incentives and we can continue to build out the ethereum stack the future of L2 is the future of ethereum and as I end every talk with the with the arbitra Mantra your chain your rules let's build ethereum together thanks