Monad Architecture and Benchmarks I Keone Hon at Pragma Denver 2025

[Music] next up I want to bring on Keon from monad to tell you about the architecture from monad and some of the benchmarks from their recent testnet launch so without further Ado please welcome K on stage all right hello everyone great to be here my name is Keon Han um from the monad foundation um today I'll be giving a little bit of a different presentation um basically I just want to very briefly show um some visualizations of the network and some examples of really bashing the network with some really computationally intensive transactions and then I'd like to invite everyone here to try it out themselves and I'll have a link to try it out so if you have your laptop feel free to um open up to the web page that I'll I'll show in a little bit just to give a very quick background on monad monad is a new evm layer one but really it's a technology project that's bringing high performance to the evm through new architectures that we think are really needed in the space um moned is building currently in private but Monet's code will be open source U prior to mainnet and more generally our goal is to introduce new architectures and Technology to help push the entire ethereum space forward um oh sorry going the wrong way all right so maybe just to start I'd like to show a quick video of um this is just a screen capture from gamon ad.com gmon ads.com and it's just a visualization of the network hopefully this works okay there we go yeah so you can just go to this website and um visualize the network right now as it stands um the monat testet went live about um what day is this today Tuesday seven days ago and it's quite exciting there's uh basically been increasing traffic um every single day um and today we saw about 250 transactions per second going through the network um this is a fully globally distributed Network um with nodes in many different countries currently um sorry I'll go back uh currently uh 57 validators with the expectation of 100 to 200 validators in consensus at Main net time that 57 is just a little bit random right now but that's that's what we went with for day one of testnet and we're now on day s Monet has 500 millisecond block times and currently has a budget of 300 million gas per second so that's 150 million gas every half second um with this fully globally distributed consensus um I think I can probably skip this slide because I just showed it all right so we'll just keep moving quickly this is a pretty familiar contract if you if you've used Foundry ever um it's just the counter so it's the default um smart contract and then I've modified it a little bit just to add some really dumb functions uh in addition so you have the increment function but then I added another function called increment many which just does a for Loop um and then I also added a couple of other functions which um push many and uh free many which just push a bunch of variables onto an array and thus um use or free Up Storage slots so this is a really dumb contract um I did this in like an hour today before my talk cuz I was procrastinating um so apologies that it's not more elaborate but I think it still shows um the overall computational throughput of of monad so with this contract here um I ran a uh well anyway I wrote a script that basically will call that contract um and so specifically it calls that increment many function with an input of 400,000 so that's just going to do a for loop from 1 to 400,000 um and it costs 141 million gas um and it landed in uh well anyway it it completed in 1.1 seconds so it basically took up a full block um and the transaction Hass is there and then there's a link to the Explorer but I wouldn't expect anyone to be able to type that in okay so again that was increment many so just showing that we can submit a really computationally expensive really dumb contract um a really dumb transaction and do a ton of work on monad um so next I did oh oh sorry that's just a screenshot of block Explorer so next I did push many so this I passed in I just tried different numbers did a binary search and I think the max was like 6700 or so um so I ran push many with 6,500 so this allocates 6,500 storage slots um if you're an evm person you know that each storage slot um like allocating a new storage slot costs about 20,000 gas um so this is quite expensive but it cost 146 million gas and again it landed within um one slot so it completed within about a actually I don't have the the output there but it completed in about a second you could check the block Explorer so basically what we have is a really high performance really high throughput environment in which anyone can build anything and worry a lot less than we do right now about gas optimization and get really high performance and get really fast finality as well because monad um has one second finality and again is fully globally distributed um but I don't want you to just believe what I'm saying I want you to try it so what I've done is I've put up a uh GitHub repo it's pretty bare Bond actually it's extremely bare bonds because I did this in like an hour um but github.com sladd developers slsp Speedtest um feel free to Fork it modify it um basically it does the stuff that I I mentioned um and then in addition I also put up a really barebones web app um m-speed test.

versel doapp um it's password gated and the password is pragma 2025 and the idea of this website is that it just allows you to do those same things that I did um but with the input your own input um so some people might be trying that out right now but here's just a screenshot of the website so it just you know allows you to choose a function name um the argument to the function and then also run a number of copies if you want to so um this basically is just a example of me calling that with 350,000 five times and it lands in the next five slots so once again you can go to mad-- test of selda app and password pragma 2025 and try it out um have fun so that that website is actually tied to an EA that has a bunch of mon tokens in it teston tokens so that you don't have to pay for gas but the other thing is that um I would love for every anyone who's here who wants to play around who wants to write their own um dumb contracts that use a ton of gas and they are really inefficient um to get some test net tokens as well so if you go to that first website there's also a link to um a Google form where you could drop your address and um uh my colleague Kevin will send you some tokens right now so that yep he's right there he'll send you some tokens right now so you can try it out as well um yeah so I'm just super excited um manad has been in development for the past three years um it's been a huge effort and we're really blessed to have the support of somebody in the crypto Community to make this happen but um it's been a huge effort to build new architecture um which I'll talk about in a second to enable this level of inefficient gas contracts that are really dumb um that can build anything all right so that was just a quick um you know link to the website hopefully you guys are trying it out right now sending Kevin your um testet token requests and he'll send make you a test net token whale right now um just just to explain a little bit about how this is possible how uh Monet has gotten to 300 million gas per second um on test net day one or day seven um with a expectation of raising that limit further to a billion gas per second by the time of main net in um a couple of months um I can give a quick summary of what we built and how this is possible so first of all just to give a quick summary monads fully B code evm equivalent so that means you could take any bik code built for the evm and redeploy it without any recompilation um it supports the Cancun Fork so um TLO T store M Copy you're supported um the same op codes costs as ethereums everything still just works the same way um the RPC is compatible with G um so you can use the same ethor get logs or ethor get balance or eth call or all that works the same way or if it anything doesn't just let us know and we'll we'll fix it ASAP um 500 millisecond block times with 1 second finality um and as I said before the gas limit is 150 million gas per block or 300 million gas per second this is the limit right now it's artificial and will continue to raise the limit as time goes on um and right now on day one of testnet there are 57 globally distributed validators with expectation of over 100 validators at the time main net um there's a link to a uh longer article which explains how this is possible and also check out the docs how is this possible um just to mention really briefly four major improvements as well as a number of other um smaller improvements there's a new database that fully supports um the Merkel Patricia Tree in which all ethereum state is stored um the Merkel tree is really powerful but also really expensive to keep updated so this is a huge undertak to build a database that natively supports the Merkel Patricia tree um additionally monad supports um op optimistic parallel execution for executing many transactions in parallel um and taking advantage of the high throughput and parallel access capabilities of this database thirdly monad introduces asynchronous execution which is decoupling consensus from execution to raise the budget for execution substantially and lastly monad introduces a new consent is called monad bft and I felt that since there wasn't really time to you know talk about all these things and um the more important thing is actually just the demo and letting people try out the demo afterward um I figured we would just only talk about one of them um asynchronous execution and the reason why I wanted to talk about it also is because um it's becoming a more um it's becoming a topic that I think is of greater and greater interest in the ethereum research Community as well um there's a eth research post about U potentially introducing asynchronous execution to ethereum as well and I'm super excited about this because um I think that it reflects our belief at the monat team that we're just introducing new technologies and we want to see them get adopted through the broader crypto ecosystem so I will explain very briefly uh how asynchronous execution works and I hope it's helpful because um this may be something that ethereum the ethereum community more broadly can considers as well so just to explain a little bit about the background most blockchains including ethereum have interleaved execution and interleaved execution means that a prerequisite to coming to consensus is each of the nodes executing the list of transactions that are in that block prior to in the case of the block proposer proposing the block or in the case of the other validators voting on a block and so I've kind of shown this in a really simplified Dumb Way um which is that consensus is like you know the orange uh rectangles in this diagram that take up most of the block time which makes sense because consensus is expensive this is nodes that are all the way around the world talking to each other uh potentially multiple rounds of communication and because execution is interleaved within consensus and because execution has to happen before consensus this really squeezes out the amount of time um for execution into a very small amount of time so when we started building monad we had this goal of um 500 millisecond block times actually our goal is originally 1 second block times and we've been able to shrink that since then we had a goal of having a block time of let say 500 milliseconds and the reality is that if you have that goal um then there consensus is very much going to take almost all of that time there's going to be very little time for execution and that's kind of what you see here blue but you can't even even see that it's a blue rectangle honestly but the blue rectangle for execution is very very small so asynchronous execution is just it's actually a really simple idea um it is the idea of moving execution into a separate swim Lane where it can run in parallel as soon as consensus completes and so you can kind of see here that consensus is happening on block n minus 2 then n minus one then n then n and so on and each time as soon as the block finalizes then execution starts on that same um set of uh that the transactions that have just been consensus upon and as you can see here this doesn't look like that big of a change right now because although we've moved the execution out it's still kind of doing the same thing um but then what you realize is that we can actually raise um the size of those blue blocks the size of that blue work that's being done substantially and so we can basically have a much bigger time budget for the execution so that's what we can see here in this last slide is that the Blue Block for execution is much much bigger than it was before under asynchronous execution it's a simple idea there's some um little wrinkles around spam prevention as well as um some wrinkles around um making sure that all the nod still arrive at the same state U by introducing the delayed Merkel route happy to answer any detail in the questions or later but um just to give a quick summary asynchronous execution massively raises the budget for execution so that ultimately a fully globally distributed blockchain with nodes all around the world can have a lot of budget for work that can be done um and that combined with several other optimizations which all stack on top of each other ultimately gets us to a billion gas per second all right so once again um the monat testet is live this is day seven now every single day we've seen a growth in usage from about 30 TPS on day one um growing almost linearly now to somewhere between 250 and 300 TPS um here are some links the chain ID is 10143 um you can try out the RPC um you can look at the Block Explorer you can check out the network information and you can check out testnet mon. XYZ um if you want to try out some fun applications that have deployed already on on day one of monit testnet and once again you could try out the demo you can um push PRS to my really crappy code base um and I would love for everyone to try to break monad and try to see how much work you can push through the network um once again really um honored to be here and excited about the future of evm um and the future of ethereum scaling through um various optimizations happy that Monet is live on testnet now thank you [Music]