Cheap, Safe, Nimble: DA and the Dream of Modularity | Sreeram Kannan - EigenLayer | Nick White

notes eth Denver 2024 is your hashtag we have lots of people watching live on Twitch at the moment hey twitch Friends YouTube videos are being posted uh almost immediately after these sessions wrap so plenty of ways to engage with this content after it takes place right here right now on the stage take away Margot thanks so much can you hear me can you hear me is this on oh there we go yes okay uh hi everyone thanks for coming here today um very briefly you know we're going to talk about da and we have some of the biggest names in the da space right now but why don't you guys introduce yourselves um and your project start with you sure uh so hi everyone uh my name is Dan I'm hit a product at ail which is an infrastructure layer designed to um help execu modular execution layers scale and interoperate and part of that is a DA layer yeah uh hi everybody I'm Shri Ram um we building a project called igen lay Igan lay is a mechanism to allow Staker ERS in ethereum to support arbitrary new decentralized protocols one of the first protocols we're building on top of it is Igan da which is an ethereum Centric data availability system we want to scale ethereum or scale rollups that are building on ethereum uh hey everyone I'm Brendan uh I work on ZK Tech and R&D at polygon uh not da directly but as Dan said earlier we're kind of the customer for for da so yeah hey everybody hey everyone I'm Nick White and the COO at Celestia labs and we have built Celestia which is the first modular blockchain Network so it's a L1 that specifically focuses on da and execution sorry and consensus and does not do any execution and so it's built specifically for a world where we want to have lots and lots of different rollups and execution layers built on top of one sort of unified scalable security layer um and we implemented a new technology called Data availability sampling which is like a a step function change in the verifiability and scalability of Da systems great well you know we've been talking we've been using a lot of jargon and I just want to make sure that everyone is like on the same page um let's start with what data availability is like what is it that these layers are trying to solve um anyone of you take it away I I'll I'll take that um data availability is a probably a a confusing term because it sounds similar to data storage but uh in fact data availability is not data storage it's about uh the publication of data it's about publishing data publicly and and making it verifiable for anyone to to check that that data was published um and this is important because if you want to verify some kind of uh you know blockchain system any kind of decentralized application you need to know what the inputs to that system were if you don't know what the inputs are or you don't know what the the state of the chain is you're not able to recreate the state then there's no way to either verify that the state is valid or progress the state uh if the the sort of block producer goes offline so data availability is really important for the safety and liveness of of blockchain systems I I'll just add a little bit to that really clear explanation uh why do we need data availability as a separate system as opposed to like actually having a common uh blockchain if you look at the scaling road map of ethereum one of the really cool things is you can off Outsource computation so you have a node a single node even that actually does the computation and submits a prove that the computation was done correctly so this is a mechanism to scale computation because now parallel thousands or hundreds of thousands of systems can be making computation claims to ethereum or any other chain or users who can then verify it but the problem is when computation is being outsourced what about the data relative to that computation what if the data inputs like Nick just pointed out or the data outputs were not made available and so by building a separate mechanism to scale the production and verification of data publication of data as well as the verification you can actually have blockchain systems that scale massively so that's that's the reason why data availability kind of abstracted from the rest of the blockchain makes sense yeah to totally agree with um everything here but um one thing that I will add is that um if you think about blockchains in general like l1's um they inherently do data availability um by having every participant in the system download entire blocks and so one thing that makes uh data availability Focus Solutions really interesting um that uh is that they allow um they allow the ability to be sure that data was published correctly but without actually downloading the downloading the data in question um and so for example Avail does that by uh doing data availability sampling as well with Val proofs um which is how an Avail light client um that is a client that doesn't actually download the body of a block only the headers um is still able to be certain that the data was published um because you don't necessarily care that uh about what the data is um you just care that the DAT that the chain is proceeding correctly and that all of the data has been available for the chain to verify the next block uh and maybe to just add one thing I I not not to get too far in the weeds but uh data availability has like a subtle but like very important difference in how it functions in optimistic and ZK systems so in an optimistic rollup uh safety is provided by the fact that anyone can generate a fraud proof so if the operator of a chain um decides to seal everyone's funds uh there's some challenge period and anyone can provide approv saying hey the transaction where you minted uh a million eth was in valid and so we're going to uh change the state of the train and so data availability for optimistic systems is really really important because if you don't have the uh input data um then you can't generate a fraud proof and so data withholding is equivalent to a safety failure because if data is withheld then the operators of an optimistic rollup can steal everyone's funds and so the uh in the ZK case it's slightly different um it's still really important to have data but in a ZK rollup the operator of a chain is providing a proof and the proof guarantees that uh the operator of the chain can't mint a million eth out of thin air um the the risk is one uh of liveness where if the operator of the chain withholds data then they can make it impossible for uh a user to withdraw uh their funds from the chain and they can halt the progression of the chain and so this uh this can lead to Ransom attacks and and and can you know is an attack Vector for xer proofs but or for ZK rollups but um it's like a really important distinction to make that the uh like there's a very substantive difference between being able to steal everyone's funds in the event of a data withholding attack and having to navigate sort of like you know the expected value of a ransom attack and and having to process Force transactions and so I I I do think it's important to note that um like this guarantee provides different property in different systems I want to throw it back to you Brendan because um you mentioned you know in your introduction that you're sort of representing the customer right and ethereum has sort of paved like pivoted its road map to this like rollup Centric age um and I wanted to have your you know take on what data availability Solutions like do for an L2 like what problems it is they're specifically trying to solve and also what are these trade-offs of having a you know external da solution versus like data from an L1 yeah so I I think the core thing to note is uh data availability is actually very expensive on ethereum and so if you look at what drives transaction fees on l2s it's uh overwhelmingly data availability the cost of posting data to ethereum and so data availability layers or systems um provide a a really compelling alternative to posting data directly to ethereum for applications that don't require complete ethereum security I I do think it's really important to note that um like if you use a different settlement layer from the layer that's providing data availability you're getting a slightly different guarantee because you can't verify that data was made available on etherum there's no way to prove like okay uh actually data in fact was made available on this other system um and so you rely on uh like the consensus of the data availability layer so so you verify that um validators on Celestia in fact uh attested that data was made available on Celestia or ail or Ian da and so I I think that this can provide a a very sufficient uh level of security for uh a ton of use cases but it's important to distinguish the guarantee that's being provided by ethereum data availability from the guarantee that's being provided by external data availability systems I I I I think it's fair to say that we need as much data availability as as possible even after dank shorting like our our total capacity for data is is I think fairly low but it's important to like distinguish the the security guarantees well you bring up Proto dang starting which is great because I wanted to get into uh you know this dunun upgrade that's happening next week in 10 days I don't can't do the math right now but um but um I guess like you know what everyone is most excited about is these like lesser fees for for for l2s um and I wanted to hear from you guys what this specifically does for like da Solutions and then on top of it you know there's been an argument that we the da Solutions are needed because of of of this data problem but on the other hand you know it's supposed to sort of solve sharding for now since charting is so far away like what happens when full sharding will eventually be implemented like what does that mean for a DA solution yeah happy to lead here um um so um I I I think that uh Proto dang charting um is a great step um it's um you know the next step in the ethereum road map uh towards increasing its block size um but my sense is that there's there's a lot of latent demand out there um essentially since the pivot to a rollup Centric uh road map you know back in 2020 um there's been a ton of work in creating new Roll-Ups and making easier to develop new Roll-Ups and we're now at the point where we have dozens maybe hundreds of rollups just already in existence um and it's super easy to it's getting very easy to spin up new ones um there are you know rollup as a service providers uh for for example that make it extremely easy to develop new ones um there's a lot of interest from developers to um encapsulate their app in a rollup and so we think that what we have now is just the tip of the iceberg um and there's there's a ton more coming um and so um um EIP 4844 I think is a great step but I think it it will you know quickly be filled up um and um you know the larger ethereum road map um is is great uh I I don't um the the Avail uh technology stack is actually um has a bunch of the same elements that uh the etherum road map um implements like we use kcg commitments um with data availability sampling um for example um and so we we believe in the technology that is being proposed for for ethereum um we already have it uh implemented for change that are you know built on a veil um and for chains that are um you know that want to settle on on ethereum like a validium for example or or or optimium um we have a ZK bridge that um that allows verification on the ethereum side that the Avail chain has come to consensus that the validator set of Avail has come to consensus um so we think that that's a you know um great way to expand like drastically expand the throughput of the system and of course every solution has to be evaluated like as Brendan was saying um you have to look carefully at what the specific trade-offs that you're making when you deploy you know one construction versus another um but um but yeah that's that's my take want yeah I I'll just add something to this so you know the there are two ways to look at a rollup we are offloading data from ethereum to a rub or we are onboarding Cloud to crypto we're in the second camp we are on onboarding Cloud to crypto what do I mean by that you look at the cloud you have services that are you know built on top of the cloud these services do not give rigid guarantees to their users whether it's WiFi ability whether it's immutable apis whether it's anti Ren seeking whether is user governance we need all these properties for end users and what is the best way for us to provide that and I think crypto and blockchain is the right answer for that and so if we want to bring all these applications into crypto one of the most important things we need is a scalable substrate and ethereum already has a scalable substrate in terms of you know liquidity you know social consensus that lots of people are interested in watching and um developer involvement and so on so can we actually take this system and add like a scaling layer to it just for data you know availability for example because that is the key bottleneck in the rollup C Centric road mapap so ethereum's like 4844 data bandwidth is tens of kilobytes per second 's launch throughput is 10 megabytes per second so you're seeing several orders of magnitude there but we will not stop there we want to keep scaling this till we actually hit Cloud scale so when we're thinking about cloud like one of the other things we need to think about is economics you know today most of the rollup or like blockchain economics are all spot prices like I have to go and B for Block space on a spot basis and if you're sharing common block space all data availability or any other system across multiple different applications one application can crowd out the entire block space suddenly this leads to bursts in prices and inability to access block space for other applications one of the things we are trying to do is to resolve this problem by actually having similar economics as the cloud for example reserved bandwidth most of the Amazon web services is reserved instances you go and have this inst just for yourself can we have data band with reserved for your own rollup or your application so that you're not contending with anybody else you're not having variable prices you have a fixed price for this entire year of horizon so this is some of the elements from the cloud some of the elements from crypto is can you pay in your own token for your you know rollups data availability service can you use your own native token off your rollup to actually also additionally secure the data availability layer these are the features that I rings that are unique and differentiated from ethereum uh I want to uh sort of touch on the original question about uh ethereum's uh da road map um and one of the core things that we realized when we were building Celestia was that there was a massive need or like latent demand essentially for data availability that was not being sort of supplied by the market because ethereum block space was just too constrained and uh since we launched in at the end of last year year uh around 10 different rollups have either migrated or deployed on Celestia uh that were formerly some of them were formerly ethereum l2s and they've seen a 100x reduction in their their costs and so teams like Manta for example have already saved around $2 million in fees and they've passed those savings on to users and they've also been able to capture some of that savings for their sequencer and they're using those funds to fund more development um and I think this is a really uh powerful Catalyst for the acceleration of more RS be launched and we're actually seeing now a new wave of demand so there's like this concept in economics which is called induced demand which is when you know the price of of some kind of service or good is high the demand for it is actually lower when you decrease the price there's actually more people come out of the woodwork basically saying I want I want more of that thing now that it's cheaper it's more useful to me or like it makes sense now for me to uh to to to build a new rollup for example and so we're seeing I think in the next few months we're going to see an acceleration of more rollups and specifically there's a lot of people interested in deploying l3s or basically taking an application that exists on a on on ethereum or some other uh monolithic L1 or some L2 and deploying their own evm or app specific chain and uh so I think we're like really just at the tip of the iceberg kind of like what um my colleague was saying over there and I think it's a it's an exciting time I think EIP 4844 is going to be good because it'll reduce the costs further but it still won't be enough at least in the near term you know there's been a lot of hype Around Da Solutions right ail has had a bunch of announcements Celestia had this main net event that happened in October November October uh airdrop was like sort of heard around the world and Igan lay you know there's been a lot going on reaking There's 7 billion dollar locked into you know TBL um and I think like a lot of folks are also have a hype around this potential airdrop that might happen for igen layer so are you worried that you know once this airdrop does happen that activity around igen layer might dissipate and users might you know go to other protocols um we are not focused on stake and tvl our main focus is on open Innovation open Innovation is what are we unleashing for developers to build on top of us that is the primary focus tvl is a short-term metric where you know users who think that there is enough rewards come to the system but the real long-term value of the system comes from the ability of open Innovation the ability that now you can have hundreds or hundreds of thousands of rollups actually building on a DA or a lot of these other services that are coming up on igen layer uh like oracles AI inference um fully homomorphic encryption uh articles that actually check what your you know https information is which is locked inside a web server these kinds of services unleash a whole new category of products that can be built on top of crypto that is what we excited about we don't focus on the tvl metric it it is a short-term metric which can fluctuate massively and you know I also want to encourage you know people who are you know staking and so on to also adopt this long-term mindset that's fair I wanted to ask sort of you know this is a new emerging space in the ethereum ecosystem um and you're solving a whole bunch of issues but what are there must be shortcomings so what are these shortcomings that maybe maybe Brendan you can take this because you know you're you're you know why would you plug into an Avail or into a Celestia or into an Igan da like what are what's still missing that um you know uh what What's missing on the on the Da Side in terms like what are the shortcomings that yeah I I mean I I referenced did earlier that uh like external da if if you're settling on ethereum which is what we're doing at polygon um external da provides a different security guarantee than using ethereum for da um but we are already planning to integrate uh or or at least offer a rollups the option to integrate with Celestia and Ian da and aale um and so I think that uh like if you look at Proto Dan sharting I I think that there's a lot of optimism that it will reduce transaction costs and I think that it will like for two reasons for because uh call data was never meant to sort of function as as a data availability uh service and uh because we're increasing the capacity for ethereum but I I would Echo what s Ram was saying earlier that um I think that like a real barrier to uh unlocking latent demand is obviously the fact that fees are too high and uh and it's pricing users out but it's also like fee volatility and the variance and not having the ability to offer a consistent user experience uh because applications can get crowded out by other sort of unrelated applications that might be on other chains or or completely um like separate use cases and so I think that uh increasing the total bandwidth and capacity for da is something that will only improve user experience and and unlock new application I also want oh sorry go for it I would just wanted to add that I I think if um I take a step back and think about like downsides of the the overall road map that we're following um this very rollup Centric uh road map brings a lot of extra computation uh and increased bandwidth but um it really increases uh fragmentation um in the overall experience both the for liquidity for the developer experience and the user experience and I I think anybody who has tried to transfer some assets from one L2 to another L2 or or God forbid to another L1 um you know you know that you've is a world of pain basically and as a developer if you're trying to develop an application that uses multiple chains at the same time for different purposes uh is a total nightmare um to to do right now and it's it's only going to get worse right because we right now we have you know a few dozens or hundreds of of rollups if you think of a world where we have thousands or hundreds of thousands um like obviously the current situation can scale um and and and that's why the Avail team has been working on on our Nexus uh layer which is a unification layer that um brings uh leverages our da so it it brings uh blockchain data to the DA layer and then um U uses ZK aggregation to uh prove the uh correctness of uh multiple rollups uh on top of that and so it's the combination of the ZK aggregation with the uh order and guarantees of Da that allow us to provide a unified experience uh across all of w three in a permissionless and Trust minimized way uh that's Brendan's prompt to talk about the a layer yeah uh yeah maybe there are elements of that that may have been inspired by polygon zag but but I I I think it's really important to distinguish uh between like the interoperability guarantees that are provided by sharing a DA layer and the interoperability guarantees that are provided by aggregating proofs and the interoperability guarantees that are provided by something like the agay which I I I think a lot of people misunderstand that it's about aggregating proofs it's not about aggregating proofs proof aggregation is a mechanism that allows to ensure safety and consistency between chains that are interoperating and so it's I don't want to hijack this panel to talk about the design of the agair but it's uh it's really important to uh to sort of emphasize that if you want interoperability you need to provide safety for chains that are uh interoperating at uh at lower latency than ethereum block finality time and and further to provide safety for chains that share uh assets and and provide fun ibility of assets um uh via a unified bridge and so I I think if you look at uh what the agir provides it's both of these things it's safety for uh fungibility of assets across Chains It's also safety that guarantees that chains can't be settled in a way that's inconsistent or that Atomic bundles can't be unbundled uh when when settlement on ethereum happens and so I I think that da is a really important security mechanism uh for individual chains but it doesn't solve interoperability on its own um I think you need uh additional infrastructure um just a couple of comments on this uh fragmentation problem the first one is I think uh people overemphasize the problem of fragmentation and the reason is we are starting from a defi oriented Locus because in in financial services liquidity is the main product and liquidity gets fragmented then it's a problem but if you've ever developed any kind of like a cloud scale thing like imagine a game most of the games have scenes each scene is run on a SE separate server you cannot run the entire game on a common synchronous composition layer so synchronous composability is actually one of the most expensive things possible and so it is neither necessary nor sufficient for large scale adoption because we're starting from a very defi oriented lens people think and overemphasize this problem lots of other places where absolutely no need for super fast composition or fragmentation is not a problem fragmentation is a solution you can have highly scalable systems uh the second point is something like Ager and I I haven't looked into the design of the uh ail aggregation system but the uh something like agare has this really powerful primitive where you can use synchronous composition on on demand so we are running on asynchronous zones and when I need to create synchrony I lock the state of these two systems and then I can do synchronous composition and then release the lock and then go back to asynchronous composition so you pay for synchronous composition on demand so I just want to point out that these Solutions also are having emerging these problems also having emerging Solutions and there are other services built on ION layer for like super fast composition using cryptoeconomic guarantees so lots of interesting stuff coming up here yeah PL plus one for asynchronous uh composition being um being like a new thing that I think will will be per pervasive throughout web 3 so far the focus has been on synchronous composability and and I think that's uh overly restrictive in terms of use cases fully agree I want to you know we have like three and a half minute I'm looking at the clock here we have three and a half minutes left and so I want to get like a big picture take sort of combine my two you know last questions um but I wanted to ask sort of we now have these three major projects Right Around Da sitting on stage here and you know polygon I mean I don't need to speak for polygon everyone knows what polygon is but you know so but I want to ask sort of what this Market is going to look like because especially on the on the layer 2 landscape like we've talked about this Brendan that there's just been an explosion of l2s I can't keep up with how many l2s are just like popping up left and right um what will the market eventually look like for uh this da ecosystem um how are things going to evolve and what can we expect from like each of your Solutions um moving forward and you know Brandon what do you sort of see from from the L2 side let's start with with NI with Nicko yeah I think I think a lot of people um have this idea that da will become a commodity um and I like strongly disagree with that I think um it will be cheap and that's actually a feature not a bug it's really good to have very very cheap scalable da and this is what Celestia for example is all about I think that's also what Ian da and a are about and um what our at least at Celestia what we believe in is um economic sustainability by economy of scale so it's about providing cheap scalable da but at massive scale because if you believe that web 3 is going to be globally adopted and we're going to be running very critical things on chain um the demand for Block space is going to be absolutely massive and so it's about um being a layer that can actually support that level of scale and even if the margin on da is very very low at that level of scale it becomes economically sustainable so that's what we believe at least yeah sorry okay I'll go for it um uh one of the things I think uh rollups unleash uh is this is a little bit of a Counterpoint is the single centralized sequencer is a super power of a rollup what it means is I can do instant confirmations not only I can do instant confirmations I think thing that I'm much more excited about is we can do subjective admission control what do I mean by that right now to Pro to access a chain you need to pay a fee fee is very non-discriminative a real user cannot be discriminated from like an me bot which just up bids the real user and you know one of the really powerful things that the internet did is to offer cheap or free usage to lot of people around the world we talk a lot about decentralization in the space having gatekeeping our systems with fees is a really bad thing what a single Central sequencer can do is to do admission control admit users some people pay fees but other peop who authenticate themselves through other you know options can be admitted into the system without a fee this is going to unlock a massive amount of innovation in crypto apps which I'm very excited about yeah yeah plus one um yeah one uh thing from from from my perspective um is that um I I do think that uh da in some sense uh will be table stakes and that a lot of value will be elsewhere um in the stack and and for us uh this idea of um you know being able to unify all of web 3 um both on the you know user experience developer experience liquidity as well as with our Fusion security um product um that brings liquidity to the a LDA I think is is our take on on this this topic Brandon I'll give you the final few seconds um yeah I I mean I think that from from our perspective like there will be Network effects that acrew around rollup ecosystems that come from the ability to interoperate but I think that having uh extreme flexibility and choice and giving users the ability and chains the ability to exploit trade-offs in how they select da is something that's really really important and I think everyone else on this stage is is kind of working towards that future all right well thank you all for this discussion and thank you to the audience and enjoy the rest of your day thank you thank you [Applause] all okay hello hello my name is Mo