Tomasz Stanczak I MEV and Scaling Ethereum I Nethermind I Pragma Singapore 2024

[Music] all right we are ready for our next talk and for this next talk I want to bring on toas from nethermind to share a bit more about what is happening in the world of me and scaling ethereum so please give tamasi a big round of applause [Applause] thank hi everyone thank you carding great to see everyone here I'm trying to find my slides all right oh wow it'll be a challenge I'll have to just look from around the corner uh so we talk about the meev and scaling ethereum and it means that we both will talk a bit about me a bit about scaling ethereum but mostly we'll speak about how me influence ethereum scaling and ethereum in general um me being minor extractable value so a bit like onchain Arbitrage but I'll not be over defining it um so I would say well me the fact that meev exists it drives the innovation in the space of uh L1 performance so the Searchers M search Searchers or arbitragers Traders onchain market makers are looking for the fastest ways of simulating different blocks fastest ways of decreasing latency for uh hitting the chain with their trades fastest way of analyzing the mles and so on and so on and so on and you know it gives an opportunity to execution client implementers to think like oh someone really wants us to make it faster someone comes and like wants to possibly pay support the teams that try to optimize uh and it's great like you suddenly start feeling like that that race that normally is outside of public goods the race between the Traders building the high high frequency trading systems uh gets to the execution client implementers and core interium development and and they think like yeah let's let's chase it um the validators themselves they want to capture more me it gives them more value to the stakers so if I'm validating operator if I'm faster in block simulations and I'm faster in attesting the blocks and I can capture a bit more like a few bips more on uh of yield on top of the state assets which means that I can win more of the market which means that I go to the execution layer Builders and say can you make it better for me can you make it more stable and faster and I'm ready to to support you and pay and so on um what I think that where rarely it comes is direct payment but some kind of reputation and caring and supporting it as public goods is very important aspect and that's great to see um mle congestions well so obviously with the with me what we can what we have is like lots of um lots of transactions they go to the um they go to the space of private order flow so they escape the mle um then you have the risks of what happens if you actually go for the mle and you keep sending more and more transactions but I think this part is more of the L2 thing so I'll talk about it a bit more later I think I forgot to remove it from here all right so let's think about the ethereum road map uh the upcoming hard Works uh on the oh I should have mentioned it's like E roadmap.com I think or.org uh so that website talks about what's coming to ethereum and soon we'll have the prag Electra fork or maybe soon is like there's this optimistic 100 days here but I think that we could could expect it's more like the q1 Q2 uh next year and there is a series of improvements that come to ethereum and among them comes the pier dust so the further data availability improvements and you may have seen a lot of discussion around what happens with L1 and the value of ethereum if l2s take all the all the flow and stop paying for the ethereum well I think that I'm always happy when I see that something that is somewhere there on ethereum road map works very well like when people start complaining that transactions are too cheap that ethereum doesn't acquire value because transactions are too cheap and too many transactions flow to L2 this kind of like three years ago I think we would say wow amazing uh and we've ped us even more of the data availability will be available for the um next surge in the um demand for the ethereum transactions through the l2s uh while we'll work on on on many of the problems around L2 interoperability uh many of them related to me facts um let's skip the the thing that I told you about like the execution clients chasing the L1 performance uh at Nine's team what we've seen recently in the last two or three months the team actually found a few fantastic optimizations to the client that led us to jump like 50 or 60% faster on the on the Block simulation time or even more like the the shortening so like in the faster it would say 200 100% faster uh and it means that we started talking about this Narrative of the Giga gas uh Giga gas meaning that gigas per second meaning that you actually can uh get the execution of the Giga gas of the evm uh gas usage within a second of block execution and compar it to the current ethereum maget where we have the target for the block size is 15 million gas per 12 seconds so it's something like a something above 1 million gas so you you're targeting something that is 1,000 times faster on ethereum uh however this is a bit of simplification we talking about the processing speed of the evm but the storage as well however in the actual real environment production environment of processing interium main not however with this speeds the storage processing and the amount of storage and the growth of the growth of the state would be quite a challenge however we address the um like the dreams of many of the l2s saying even if you cannot handle that on L1 on L1 main net then the l2s will be very happy to take it and we see the massive improvements on the processing times for the blocks in neine uh these are some examples of the uh nowadays stability of of Nom mind and possibly many other of the execution clients would show you similarly improving charts um and also the average times latency to process the block and the how much of the meag gas per second you will see so those little BL blue dots and on the right side the scale showing you something between zero and 1. 2,000 megag per second so uh many of those numbers they actually reach for single block the gig per second but here it's a on a reasonable strong machine but the machine not in the data center but the machine belonging to one of our our Engineers uh it's producing something that average 700 megag per second now the solo staking performance so not not only we are able to process blocks faster and better with lower latencies when we execute them on powerful machines but it also means that uh the requirements from the nodes themselves are lower and lower for operating a single validator and here I have example of the gum Network testing execution of 100 valid validators actually went they went further and run 200 validators on a single Raspberry Pi 5 machine so suddenly solo staking is not that expensive as it used to be and it's interesting because four or five years ago we were saying that the ethereum state will continue growing it'll be harder and harder to do the solo staking it will be harder to operate a single machine but the improvements on the logarithmic side and the improvements from the execution layer clients uh and improvements to evm allow us to actually run it on a smaller smaller Hardware less expensive Hardware operating the validators at home and I think we start to see the optic in the solo staking numbers on ethereum also thanks to the uh incoming Technologies use Technologies from naido and others on DVT uh so it would be like SSV Network and Obel are think the leading Solutions but there are a few more and uh and Lio is adopting that and uh this least to more more solar stakers and you see also like rocket pool improving a lot so where else there is me so me affects the L2 decentralization design so um maybe if you seen oh sorry I thought it was the next slide this one uh if you seen metallic writing recently look from next year I stopped talking about any l2s that are no longer uh that are not yet supporting the the proper security measures I'm not properly uh securing the assets that go to L2 as it should be in the rollup so really don't call themselves let's let them not call themselves rollups if they are not executing all the all the properties all the assumptions of a rollup and because of that people think about how to decentralize l2s and when you start thinking about how to decentralize l2s you have to ask lots of questions that normally you would ask for L1 because now the lots of trading activity moves to L2 so you're asking questions about how me will affect our system if we CZ versus not C we decentralized versus not uh so think about operators like a single sequencer operated by the L2 Foundation uh L2 chain and they think like the reputation at stake of the sequencer that comes from the foundation or come from the uh from the core engineering team it's it's too high to really play the me games um but if you want to decentralize it or even more make it permissionless then you have to start allowing the other operators to come into the game and the operators may be actually very happy to to start thinking well why why wouldn't I play the m game which means that the cord Engineers will have to think about me all the time as the main driver or the decision making in how to design permission lessness the centralization the sequencing uh mle designs for l2s now the it's not like suddenly we have a measure that forces out to decentralization previously we didn't and just before me when we were thinking about the centralization we're thinking more something like what if a what if a nation state attacks the chain what if we are censored because someone tries to censor the political speech or activity of some of the rebels in the country so some somehow that's what I was thinking about when I was thinking about the centralization years ago and I was joining blockchain I thought like blockchain protects us on the political side of the centralization but me said that look you have something much simpler for many people that do with blockchain you can simply start quantifying me as a like the centralization as the amount of money that you can extract in the short term by by censoring others and all of that affects that decentralization designs speed up a bit okay so like U one and a half year ago or so we started this exploration of various design on l2s how they can work in the format of based proposer so the proposer where the L1 And1 block proposers are proposing the blocks on L2 so like it's in the case of Tao rollup uh then we looked at various like shared proposal setups PBS in the context of PBS 401 PBS on all2 PBS in the arbitrum setup with the fcfs so how it can be proposed there PBS in in ZK rollups in optimistic rollups and so on so it started massive exploration of how we can push the l2s to the centralization because we wanted to address the questions of the Searchers on how they will be making money and will there be me left on the table a in those l2s so once again M was driving much of the research and this is research that's very often sponsored uh by highly well capitalized entities within the like market makers Searchers and financial institutions um and yeah when we talk about scaling ethereum and scaling ethereum via l2s look at this nice list of the l2s on top of ethereum uh the assets uh sorry total tvl total total value locked on all of those l2s uh growing obviously maybe after the last year it's it's it stays more or less stable but comparing to two three years ago it's growing uh substantially and we have a lot of activity on the major l2s um whatever you see on the top I'll still uh say that pay attention to starnet it's it's biased but yeah there's fantastic execution from from arbitrium base op other chains we had Sundy talking about scroll uh and we're doing everything uh great for stet to keep jumping forward the this is that least that uh vitalic is complaining about look the l2b says we we're no longer pretending let's find some metrics that will look very very red for all the l2s and uh drive that scaling drive that scaling in a proper way uh so these are no longer centralized sequencers and actually if you want to do the escape hatch you can escape if you want to ese ese uh sensorship you can escape censorship it doesn't look best so far but there's lots of promise that now when L2 is deployed they launch tokens they're happy to think of all right let's just make it so nobody complains anymore uh what else comes to the L2 like when we think about what uh Justin t pushes a lot with the prec confirmations and sequencing calls so prec confirmations leading to the sorry the the base rollups proposal as the uh most most L1 native approach to to l2's most aligned approach to l2's based rollups and face rollups answering question like look we produce the L2 blocks only at the rate of the L1 block every 12 seconds how we can address it will address it with prec confirmation so you actually work in a different way somebody who sends transaction sends that to the to the sequencer or gate where some other operator they send back information I pre-confirmation of your will be included in the next block so almost like uh creating chunks of the blocks that are confirming and are slash those are like slash promises and then if the my transaction is preconiran-food.com all of this would lead to the probably more powerful based rollup designs uh and there is a working group in etherum around commit boost comit boost designed as a public good with the funding from uh leading uh leading protocols on ethereum and with the uh team and actually uh collaboration of multiple teams uh expanding the ideas from M boost or maybe like working around me boost comit boost building pre-confirmation to the PB s design once again protocol driven by the uh by the me requirements me driving the research and coordination of the protocol improvements um what else M affects well it does affect behavior of the validators in the L1 main net so first of all it affects Geographic decentralization uh and validators potentially aim at collocation with the signal so what I say is like when you want to trade then IDE ideally the signal to the trade appears in a place in the world where you very quickly capture it process it through your other Market data process it through your algorith and then you send it ideally to the in a PBS mode to the local relay and ideally the validator is also local to you so you remove all the latencies of going around the globe uh but you can think of the situations when the trading signal appears in New York versus Tokyo versus Europe and the operator of the me searching appear operates from from a different country and they relays in a different country and so on so actually validators start to look at the setups uh and the tooling and the protocols that can speed up delivery of the signal speed up delivery of the blocks to the uh to the validators and having proper distribution uh and then it may affect the geographical decentralization because the might be a more of the push for the better me payoff payouts rather than the security from anti- slashing from removing the coll uh sorry removing the correlations because the correlations lead to correlated slashings uh and and those are one some of the biggest risks within the operation of the validators also the timing games is like delaying publishing the block just to extract more Me From the Block once again you have the protocol that assumes that the block will be sent distributed without the delay to collect dat stations but what validators would do often they would Delight that block by one two 3 seconds just to extract more me and this becomes quite visible in the uh in the network uh further me drives are some of the EIP like the D protocol changes on the L1 so 7732 is the enshrined PBS so PBS that at the moment is outside of protocol as a as a site car is potentially coming to the uh to the main protocol design and lots of conversation on this topic there is also the breakout sessions that you can follow on YouTube uh geographical decentralization so here you just to to visualize that here is where majority of the validators are operating from so you see quite a lot of them in Europe and then in us uh much less in Asia but I think we put a lot of effort into educating in Asia and in other countries of like how to operate validators there how to extract M locally however we will always be affected about with the value of the signal value of the economic signal for trading uh because very often you have this intertwined traditional finance and onchain markets so the signal appears in the traditional Finance markets that drives later the trading onchain so the signal appears in the major financial markets in the world which would be New York London Singapore but actually probably like 70 80% it's all in us uh the timing g games you see here here the behaviors of different validators the red color is when uh the blocks are being delayed and published with a significant delay and talk let's talk also about two more things so the M drives the sensorship resistance designs once again for this reason that I mentioned to you it's not really censorship in the maybe political sense but censorship in a sense of I compete with others for the money extraction so I censor them I may pay for not including the transactions so I can include the transactions that will extract the value uh so M drives the new ideas and massive research in the space of the uh censorship resistance implementations and fossil is the fork Choice driven inclusion lists proposal that is now discussed in the last uh probably four weeks or so you also have multiple concurrent proposals that you see as a proposal on salana and now driven also to ethereum as one of the solutions and well I mentioned centralization I'll show you two more uh boards quickly uh so the big problem nowadays with PBS is it's centralizing with the blog Builders a few major blog Builders producing a lot of blocks so me does also Drive the centralization of where the blocks are produced because the value is extracted the better and better Builders are constructing better blocks and recycle that and then win more of the block space Also multiblock multiblock M discussions are Rush a bit now uh but it's interesting topic finally we see some examples of multiblock multiblock meev which were not visible before um I encourage you to also wi read the who wins ethereum block building auctions and why uh great uh great analysis of all the behavior of the Searchers and the blog Builders and also categorization of various aspects of Meb thank you so much for listening and uh I hope you look at scaling ethereum scaling with me through me because of me thanks a lot [Music]