Mega Labs - How megaETH will scale payment systems

well some of you may have heard of Mega some of you may have not it's been along Twitter a lot these days I think if there's one thing you need to take away from this presentation is we are taking an extreme approach to what it means to scale a network we are here to fundamentally unlock new use cases that have not existed before and in my humble opinion we are one of the stronger performance unlocks for crypto in recent history one sec okay A bit about our team we have a pretty strong domain expertise when it comes to low latency computation high performance networking and a general Vibe spirit and technology behind blockchains right why are we actually here elong studied uh High frut pardon me elong actually studied low latency computation and Stanford for his PhD while lay studied High frut networking for his PhD at MIT for those of you in to know actually was one of the co-authors of the famous data availability white paper with Sham from aan Lair shya has been a consensus for six years where she led Global Business Development overlooking metamask Linea and infura I'm the you know local djen I've done a wide range of things from investing in infrastructure protocols Adventure trading liquids at a prop shop and actually being an operator from organizations ranging from Dows to uh companies like consent ensus our fundamental thesis is similar to many that layer ones fundamentally cannot scale proportionally to the world compute the way we receive these conclusions is however somewhat different we think that performance and decentralization are fundamentally quite incompatible I.E there's large tradeoffs that you need to make to be decentralized you can see this in the market today if you're is extremely decentralized right and because of that it's not very performant salana is a bit less decentralized right the nodes are beefier meaning it's way more performant and it's just a more active blockchain that can handle more things um that's like a pretty unfortunate Truth for you to be more scalable on the L1 your nodes need to be beefier it's a conversation that a lot of high profile ones don't actually want to have you can think of this best by an analogy right let's say there a thousand students in your class and you're a teacher there's some really bright kids and some kids that are less bright you have to go at the same speed as the slowest kids to make sure they understand the content blockchains are similar today they fundamentally throttle so that the slowest participant is not dropped this is well known as the straggler problem and that doesn't mean that there aren't optimizations you can do at the L1 right mon out famously is paralyzing the evm our feis though is when we come to this question how do we scale to World compute it's fundamentally Impossible on the L1 and that's because we believe that even a modum of decentralization right results in massive performance degradation our Focus however is to focus on the last mile scaling bottleneck and that means execution we believe that you know we are performance first there should be a strong division of labor approach when it comes to layer ones and layer twos copy pasted that from met alic uh blog post from 3 days ago by the way and the way we do that I'll walk into it end result is 100,000 transactions per second one millisecond latency we do this using two techniques the first is something called node specialization what we do is we make sure that every single node on our network is heterogeneous heterogeneous means that the node does a singular task while on traditional blockchain architectures nodes do everything right each node that's every single task that's fundamentally a bit of a problem right because there's a mismatch between the tasks in the blockchain execution sequencing validation the cohorts of nodes that will run them and the hardware and software capabilities of those nodes right what we do instead is we say Hey what if we looked at the task itself and figured out kind of like reverse engineer it what's the ideal node architecture so for execution we should probably have one node that does that right that's extremely beefy for validation we should have as many nodes to do that because you know you're validating Network by doing this we were able to have specific hardware and software configurations for our architecture and I think that's best visualized by like our sequencer right highly highly performant uh we're able to push a lot of blocks right 100,000 transactions per second subc latency by having a really really performant node here you're going to see our three different notes compared to just a traditional Market aerium salon and aptose you can see for Block production it's really really intense 100 core CPU four memory terabytes Network 10 G gigabytes you know it's pretty pretty strong and then you can see that's not the case for everyone else but they can't because you still need to make sure that the the chain is decentralized on L1 right you don't want to make it too expensive however when it comes to validation we're also much much lower than the most decentralized l 1 which is aium so in some ways we're actually the most decentralized blockchain when it comes to validation yeah so you know we have this extremely performing block production but that's not enough right one performant piece of the puzzle will result in bottlenecks chowing up in every single other piece of the puzzle so our answer is we actually need to optimize the entire pie here's a basic architecture of how Mega E looks on the sequencer as mentioned before we have a highly high-end ser 100 Core 1 TB Ram 10 GB Network right but there's additional pieces here too on the sequencer we're able to do because it's so strong a variety of optimizations we do in-memory Computing jit compiling parallelization on the pro architecture effectively what we do is we have a prover Network where each prover node does a subset of the transactions of the sequencer this architecture is similar or also known as stateless validation it allows us to decouple the node requirements of this that are associated traditionally with the sequencer and make the node extremely cheap right so the goal is for you to eventually be able to run the validation node on a phone I mean this is a iPhone's a bit overpriced but you kind of get what I mean uh traditionally right the prover network will then submit proofs to the full node and now we're on to the full node right so the full node also needs further optimizations if you have such a performance sequencer right the full node will have a variety of problems traditionally the sequence the full note has had a hard time maintaining the state all of this information especially for a uh you know architecture as performant as ours so what's the solution here we've actually Rewritten the state try so that's point a point B is that we've actually all of our state Deltas State dips are compressed so by doing this we're able to optimize the variety of bottlenecks that show up when you have a highly performant block production as I mentioned before the results pretty exciting right we have built the first real-time blockchain 100K transactions per second submillisecond latency we like to call this instant responsiveness and 1 millisecond block times I think an important conversation to have right now is about like decentralization right we've basically told you we're proudly centralizing block production it's important to note here that this is what we believe is the role of the L2 is you can make these tradeoffs that you could not make on L1 so we effectively are able to bake a bunch of security measures that are inherent to l2s to justify the tradeoff we've made a we have proofs to ensure that the sequencer is following the rules initially we're using optimistic uh architecture right and the reason for this is truth be told ZK is still pretty expensive obviously long term though that's the plan secondly you have the guaranteed rights of effectively Force exiting if you don't enjoy or you don't like the behavior on an L2 you're able to go back to the L1 it's not ideal but it works right it's a really strong uh fail case mechanism finally you have freedom of choice meaning yes you have one sequencer that's centralized but that doesn't mean you can't rotate that sequencer that's the unlock here by having one powerful sequencer you're able to have this level of latency you're able to have that level of performance but you can have a rotation of who is the sequencer so there are free ways we're able to inherent the security and decentralization of ethereum now why have I come here and talk to you guys about infrastructure when this is clearly a stable coin Summit well because weing there's a massive unlock for stable coins and payment systems this is in my opinion a very simple flow of what transactions look like today it's quite friction full to on-ramp transfer and off-ramp it's at least free transactions involving an on-ramp vendor including client and an off-ramp vendor Tron and salana have meaningfully lowered the cost associated with Payment Systems as a result you can see that the PayPal salana integration when strip did their cohort you know a couple years ago almost all of the projects which were part of their cohort were on salana because you can actually build more interesting uses of payment on salana because of the costs and Tron has found insane product Market fit in cross border transfers right all of this is because they've been able to substantially lower the cost of Serv s of these transactions We Believe Mega can take this one magnitude further by bringing the cost effectively to zero it makes it highly valuable for gasmaster models to enable new to Consumer businesses yeah so here's three examples right in a world where USD transfer goes to zero we believe that there's a lot of unlocks Associated around to Consumer applications there are a couple examples are crossb transfers I'm azer bani like I'm telling you would be quite useful there are Entertainment app payment mechanisms so everyone got into crypto of the IDE of hey let's try and like decentralize existing like middleman businesses you can now create really interesting trickle streaming mechanisms and finally like factoring businesses as well it's interesting right I think one strong rebuttal can be like well the cost is already so low for a transaction on salana we believe that the difference between small and effectively zero is massive and you can think of this similar to web 2 where the cost of data kept decreasing by a factor of every single year a little bit about how we're going a a little about how we're going a pro I'm not English is not my first language a little about how we are going about ecosystem development it's very targeted right we think that this is a phenomenal unlock and it's very very exciting and obviously I have ideas our team has ideas but some of you might have way better ideas so we work very very closely with brilliant talented builders Founders devs on creating applications and use cases that could never exist before the biggest paino with web 2 for web 2 users who have pre is the fact that like you have to wait latency sucks there's no more latency issues with mega what can we build what kind of experiences can we create that's why I'm here today to talk to you guys about these if there's something you want to build definitely reach out and we can work together we're on a goal to make a fre room great again and and yeah this is Mega def Which hat Good Vibes would love to chat with you guys after the call uh after this talk awesome that's it for on my end guys