Modular Layer2's by Arjun Kalsy

hi everyone welcome to the main stage I'm Steph Bell the next talk we have is modular layer twos which will be a discussion about l2s and modularity and Cooper is going to come up he'll be the moderator here so give them a hand he's going to do some introductions everybody we have a really exciting panel and a topic that I think is really relevant for the ecosystem today we know there's a lot of l2s launching and also the discussion about modularity and how we can improve the fundamental architecture the panel today we have from mantle Network our head of ecosystem Arjun Kazi a software engineer from mantle AFK and founder of eigenlayer sriram so let's start by just teeing up kind of where the ecosystem is at today Arjun what do you think are the key challenges that layer twos and scaling Solutions today face so um all right wow that's that's loud um so if you look at look at what's happening right now so the L2 model to an extent has has kind of uh let's say product Market fit there's lots of l2s launching uh the many different types of chains we're also going to be seeing ZK VM chains you know going into possibly some sort of baby Alpha main net depending on which team you ask in in the weeks to come um but but I think what has to also happen is there has to be a some sort of evolution on the basic chain architecture and design itself which is kind of what we were trying to solve at mantle so we realized that you can build a chain but if you build it with a modular architecture in mind that's when the chain becomes a lot more scalable it becomes upgradable and at the same time you can continue to sort of swap out components and continue to sort of upgrade that particular piece of infrastructure with new capabilities so for example if you have a modular chain you can do data availability or you could have different types of fraud proofs or other kinds of let's say features which you could add to the chain but this is only possible if the test bed or the core architecture supports this kind of development so this is kind of why we came up with the modular design and you know we've been working with with SRI Rams and his team for quite some time um and yeah so very excited about this sort of shift in thinking towards modular architecture because we we have belief in this design choice of it being sort of the model for the future of ethereum L2 chain design that's great and of course you alluded to our partnership with eigenlayer I think this is a great place to ask freerom if you describe the role of data availability in layer 2 scaling and why is it important yeah very excited to be here thank you so much for inviting me to the panel um the so when you look at data availability uh what what is what is it trying to do right the the idea is that whenever you want to run a layer 2 solution you're offloading computation and ex you know memory management to an off-chain system so when you're offloading the computation to the off-chain system and submitting proofs that these systems are working correctly we also need to have a place to publish the inputs to the computation and if you publish the inputs to the computation anybody can verify that the computation was indeed done correctly or continue on your system even if the sequencer goes down so both for liveness and for safety we need to have a place where we publish data the inputs to the computation in a very secure Manner and right now most of the layer 2 Solutions are publishing data directly to ethereum and while this is possibly the safest solution you know because you're publishing it natively on the ethereum you know call data this is can be expensive this is expensive because the underlying system is non-scalable as of today and so when we think about data availability the question is where do we publish the data so that the inputs to the computation so that anybody who wants to have access to it can actually access it and when we want to think about the data availability problem the way we think about it is how do we provide scalable data availability how do we publish data in a way that you can actually have a large decentralized Network securing it but at the same time every node doesn't have to incur the cost of replicated Network bandwidth and replicated storage everybody is not doing redundant work that's not a way to scale so that's a high level overview of why data availability fits in with layer tools and with that understanding I think uh we should better understand how mantle and eigenally are working together Arjun can you tell us how mantle is addressing data availability with layer 2 scaling well 100 right so this is kind of where our collaboration with the you know with eigen layer comes to you know comes with four because data availability is still in development many other teams are trying to sort of figure this out in our research we realized that you know sort of um eigen sriram's team has has sort of uh you know I had a very sort of you could say mature economic model with which how they wanted to solve this problem with E3 staking and how this approach can solve data variability but also become a layer of decentralized trust right which has more capability right so we kind of felt that we should go over the design which has a long-term roadmap sort of built into the design itself where you can do data availability but then you can also do a ton of other things um and I think if AFK you can maybe describe how the integration like sort of looks um you know we could maybe give some more details yeah so we currently have a pull request that everyone should look at on the repo as well that has a preliminary eigenda Network sort of Spun up so I'm not sure how familiar folks are with like the architecture of eigen layer at the moment but they're sort of separate products there's eigen da which can then like roll but which is a product used by rollups which could then roll up their data and post like the main eigenlayer chain so what we sort of have right now is a dual staking model where our data availability nodes are able to take both bit and eth we then take like the call data that typically is posted to L1 and the ovm V2 CTC contract and we sort of divvied up chunk it and we cast it to these nodes to be stored on that Network I actually think that that will be live in maybe two three weeks from now so I'd keep an eye out for that pretty exciting yeah right and and why this is interesting um you know from from a model perspective is that when you restake these assets right so let's say you restake eat or you sort of stake our token let's say right for our particular let's say chain infrastructure the yield on this taking is utility based yield right you see this when you're staking crypto on a decentralized trust infrastructure the yield is actually coming from actual utility which you're providing to the network right you're improving something and and this is the kind of yield which I mean this is how crypto assets should be used right to provide utility to solve problems versus you know versus how what we've seen is um collateral based yield where you essentially stake your token it becomes collateral in some some protocol you get your 10 000 APR like that's a zero-sum game right if you want yield it should come from actual utility and that's kind of what I really like about like eigen's model where you use crypto to improve things and and that's how I think it should be used uh maybe I can add a little bit of context on the economic model so just to give a high level of your eigen layer is a mechanism by which ethereum stakers can provide other services so this fits in very nicely with the mantle world view which is that you're going to have these many modules and you know it's not about committing to one particular module right now but over time as Innovation comes in you can continuously integrate these modules into the architecture that's exactly our world view as well and all we are doing is to be like a decentralized trust Marketplace to supply trust to these modules be it I want to secure data availability I want to run decentralized sequencing in the future you you know I want to run a super fast settlement you know I want to write data and you know if you want to write data to ethereum there is a certain cost and there's a certain latency can I get it done in like one second and confirm to the users so there are various levels of utility that decentralized trust can be leveraged in the modular ecosystem and these are some of the dimensions we are engaging with mantle on on the core thing that we've already started working and we're very excited that mantle has a chosen eigenlayer as a partner to build I you know integrate with eigen da the one very interesting thing about Roll-Ups is ah the one problem and I'll say how we solve it and the problem is that when you're running a roll up you have very uncertain data availability costs so the costs of data availability which is the dominant cost of running a roll up are very uncertain so you don't know how much it is going to be because it's based on congestion pricing so as you know suddenly right now in the bad market like you know prices are low and everybody's happy and then suddenly what happens is when the network utility goes closer to the network capacity then the prices start ramping up in a way that is highly unpredictable imagine you want to run an airline and you have no control on the jet fuel prices you have no way to hedge it you have no way to buy it this is a very unstable way to run a chain because you cannot guarantee your users what the price Dynamics are going to be on your system so one thing we thought about a lot in building eigenva is how do you make sure that when a roll-up like mantle wants to think about integrating with something like India how do you give them an assurance that you have a certain amount of bandwidth that you can use you can preserve it so the idea of congestion pricing comes from a scarcity mindset like there's not enough so there'll be congestion and if there is congestion we need to pay for it and of course you have to but when there is abundant block space when there is abundant data availability bandwidth you cannot price for congestion because there is no congestion you have to price in the way that clouds think about like you know an AWS for example how do they think about pricing is you can reserve an instance and you can use it for yourself and this gives a superpower to the Roll-Up a roll-up like mantle when they start using eigen da what happens is they have price certainty so that they can tell their users that yeah you know till this transaction this is how we're gonna price we know exactly how much it's going to cost us and this is I think the way to scale crypto to the next billion users we talk about this a lot but we need to understand that the under underlying rails need to have that scalability yeah of course and one of the most interesting parts of working together is kind of figuring out how we can enter this new paradigm of pricing giving users more certainty and as a role that better understanding our own economics so there's a lot of benefits here to working together uh SRI Ram do you want to go deeper into kind of the eigenlayer stack AFK alluded to some of the components that need to work together nobody knows better than you uh what how they work together and what the benefits are so when you think of something like the Mantle's modular stack right you have at the base uh you need to write your settlement somewhere so there is a settlement layer where you're writing the proofs that you've done your computation correctly and these could be optimistic or ZK that's one of the things we like about the mantle architecture is keeping itself open to the evolution in the space because one thing we can all be sure is that the technology which is best today will not be the best tomorrow so having so that's at the base like how do you do like proofs then on the top you have something like a data availability layer where you have to write the data and again in this layer for example uh the eigenlayer can can be useful is by providing data availability to ethereum Risk Takers and having an architecture which is scalable uh I I mentioned the settlement and settlement one of the issues with ethereum is the latency of finality you know ethereum takes 12 seconds to produce blood dogs and 12 minutes to create finalized blocks and when you want to assure your users that your transaction's gone in with a much higher economic certainty would use a layer like eigen layer where you can actually prove that you've done it correctly to a restaked nodes who put their collateral up to certify that that was done correctly so that's about settlement and then you move one layer up you have a sequencing so how are you ordering transactions and one of the big things with ordering transactions is there is no way to prove that you ordered all the transactions correctly and this is true for ethereum this is true for layer 2 Roll-Ups so having a single node trying to order transactions is not necessarily the optimal way because when you run complex Financial or gaming applications like mantle is intending to do the problem that you run into is how much Mev or like what is the censorship resistance that you can offer to the users and having a more decentralized way of doing it again something like eigen layer can play a role there uh there's also other things like Fast bridging and so on where you know we can have a good partnership and so we've spoken about a lot of these benefits I'd like to hear from AFK if you could narratively tell people how they should think about the benefits of modularity in a way that they can remember and take home with so honestly uh about a year or so ago a couple years ago everyone was just trying to make these monolithic elements that were all sort of competing with each other and vampires are attacking each other and everything like that but I think that with the sort of new modular narrative we can sort of come together and pull together like the best pieces of each um sort of part of the infrastructure stack so like for example mantle for our execution layer we're just using ovmv too right we're combining that with um right now eigen da looking into maybe working with other partners for fraud proofs specular and that sort of like opens up our our engineering team to trying to make as many improvements as possible to the dev developer user experience on our platform and like making instead of having to focus on you know putting this whole working and developing on this entire vertical if that makes sense so I think that um more and more chains are going to or l2s are going to start um leading in towards this we see this with the op Collective we see this with how like coinbase now instead of you know going like we're going to build our own L1 we're going to build our own L2 we're just going to leverage the existing technology and work together with them co-develop start giving back and I think that this sort of hyper collaborative um future is really what will accelerate the adoption and the acceleration of innovation within technology itself um yeah no definitely I mean the future is innovation through collaboration versus you know confrontation and I think if everybody in the ecosystem collaborates more and works together then we'll get to the promised land faster right we'll get to master adoption much much faster so we are committed to this model and and we feel that everyone else should also follow this model and I know collaborate together and sort of achieve these results absolutely I feel like something that people should take home with them is that one of the benefits of modularity and unbundling the traditional monolithic stack is it creates a lot more opportunity for synergies across the ecosystem which benefits everybody because competition does at times create waste but when we can partner together and work together it really ensures that we take the step forward as an ecosystem AFK or Arjun do you want to speak more to Mantle's specific architecture and how it improves upon existing layer 2 Solutions so I can talk about some of the stuff that we've been working on um on top of the existing like ovm 2.0 stack we currently have been experimenting with decentralizing the sequencer which I think is something that's a bit exciting we're using a bit of a naive model at the moment you can actually take a look in our docs but it's sort of like a round robin model like bft weight-based staking with like a scheduler so each of our like block Builder nodes sort of take turns either being the scheduler who will choose the next node based off to either some election protocol or like weight-based shaking and the node that actually constructs a block that is a sequencer the rest of the nodes sort of become followers if that makes sense now this model right now is a bit naive and there's like edge cases that may not be handled but we're like determined to like continue iterating um over it and in order to see like what we can do to improve it further and we're really open to working with other teams who have also been working on decentralized sequencer models to see how we can sort of like collaborate and like pull it back into the stacking make it available for everyone to be able to use some other stuff we've been experimenting with is we've been sort of like co-developing fraud proofs with specular a small company out of a big Don Chong's lab in Berkeley again like I think that we could have gone our own route and been like you know we're going to make our own fraud proofs and that's gone really well for every team that has tried to do so I believe so I think that working with a team that has the expertise lending engineering help and helping them like develop like a proof of concept helping them sort of like iterate and test through it together is again what's going to get us the farthest yeah yeah and I just wanted to also just take a quick minute to talk about how we've used multi-party computation right so this is like a very unique way of solving the optimistic roll-up problem right of seven day exit periods right so typically all optimistic Roll-Ups have a seven day exit period and then there's a reason for that right so it's a fraud proof it's optimistic in nature so when you post for example the data and the proof on on L1 there's a seven day window for anyone to prove a fraud right so if you think about it what you actually need to do is get people to prove the fraud faster because if you could do that maybe you don't have to wait seven days and this is kind of where the multi-party computation system comes into play so we've used the same MPC style which many wallets use specifically for example if you look at Zengo uses um multi-body computation with a threshold signature scheme we've used the same model created a separate set of nodes called multi-party computation nodes and there's also another set of roads called verifier nodes and this is like super interesting so whenever the sequencer creates a proof the multi-party computation node Network which is a separate node Network and also decentralized creates the same proof right and then the verify nodes sort of check both those proofs so if the sequencer and the MPC notes the proofs match then the proof is considered to be correct right and depending on the decentralization of the MPC node Network and the verifier nodes you can almost prove fraud immediately right so you're at seven day exit period could become three days could become one day or possibly even a few minutes depending on how you set up this network so so we've kind of done many interesting things you know I encourage you guys to look at our GitHub repo see our code everything is open source but we've tried to take existing optimistic role of designs right and continuously sort of iterate and improve on them and we'll continue to do that and and bring like new technology and new techniques to the force so yeah and of course some parts of this design are very intended to pass off gas savings and cost savings to users sriram could you talk a bit about how this architecture contributes to that yeah um like like I was saying one of the most important things is having price certainty not just price certainty price certainty at a low price point and a high throughput is what we're looking for seems like you're asking for too much but that's not really the case because there are improvements in the distributed systems architecture and actually work done by EF which is really you know pushing the frontier and this the core idea of how to get hyperscale data availability dog sharding is a kind of a major milestone in this uh in in this Evolution and what we've done is like you know mantle was saying building on existing great ideas you know we took the Don sharing architecture and basically made it uh come usable with the eigen layer core decentralized trust model in a way that actually it scales even better than what was originally anticipated so instead of right now on ethereum we're having 83 kilobytes per second throughput on eigen da we have a 10 megabytes per second uh throughput and we have a roadmap to scale this even further so once data becomes abundant and you're not it's not just a way of making data abundant by asking all the nodes to do a lot of work it is by asking the nodes to a lot of nodes to do a small amount of work that's how we get uh that's how we basically get the cost savings and this is passed on back to the mantle roll up and to the users so we have one final question here which uh contributes to the broader picture of mantle and Agony are together within the bit doubt ecosystem there's many parties RJ could you briefly tell us how this ecosystem contributes to the overall vision of what we want to do within web3 100 right so uh will set up with the sole objective of building web free technology for master adoption right so if you look at all the different pillars of mass adoption when you look at Master adoption like a problem and you start working backwards you need to solve obviously infrastructure you you need to solve research you need to solve education right and you also maybe need to look at sort of other structures which bring in a huge number of users right so if you look at what beta is doing right now is sort of incubating mantle right which is next Generation L2 infrastructure built for super high scalability beta was also invested in game seven because gaming has the potential to bring in a huge number of users there's also ZK Dao which does research in the ZK space this Edge which is solving web free education so all the different pillars you need for web3 master adoption is is what beta is compared to building right and then it operates like a Dao so it's completely decentralized right which is interesting for us because chains typically all the chains in the world are super centralized and become decentralized or a Tau at some point in the future we're already there that's our starting point and and this is where we will you know continue to build and and because we are a Dao we're committed to building and collaboration right this is because this is how Taos also scale dowels don't scale vertically like a tech stack right they scale horizontally with the size of the community so the more Partners we work with I think the better the outcome will be for the battery space as a whole so that's beta's Mission and that automatically becomes Mantle's Mission it's which kind of you know gives you a sense of the kind of design choices we made and our philosophy uh towards Building Technology so yeah thank you so much for that and that concludes this panel I'd like to invite you to a warm Round of Applause for our panelists thank you all [Music]