Koen van Marrewijk, ZKV: Decentralising validator sets | ETHSofia

[Music] I'm the devops engineer for zkv a quick blog for zkv we are proof of stake um validator we do other things among uh ZK approving shared sequencing um and also invest in ZK projects but also any other kind of um kind project in the blockchain space that we feel excited about cool so decentralized Val their sets um well if you think about it do we really need to um do we really need to decentralize further right if we um look at kind of the reasons for decentralizing it's U reducing trust um deterring the ability to control a certain Network um which might have you know hundreds of millions of dollars of of um liquidity on it um stability um you might not you know we care about certain stability in networks certain networks um have been down in the past un not sure if you've ever had issues trying to send transactions and it just not working um so it's cool to have a level of um stability in networks um resource distribution so like if we look at the amount of compute that's on the networks um it's quite important to have a certain level of computes which one validator or one um node might not have um available um so that you can kind of collaborate with other nodes in the network to um still do certain things that's kind of what ZK proving to an extent does depending on the networks um so how do how do we kind of decentralize now right so there's many validators in certain networks so there's um networks like um ethereum that have hundreds of thousands of nodes but there's also networks um mainly in the Cal ecosystem that only have 100 for example um so it depends on the network but usually as many validators um in ethereum especially these validators are actually so there might be hundreds of thousands of validators but there's not that many nodes because you can actually run 30 50 100 validators on one physical machine meaning that you know every other every every next Val data that you add is not going to add much decentralization to these um to these networks um a lot of staking currently happens through um major um staking protocols like Lio um also not necessarily helping decentralization um the moment that you stake with them um you know you get yield and you get tokens to them um but actual decentralization doesn't really um uh you don't really contribute to decentralization of these networks um at home saking has been talked about a lot in the last few well in the last year or so um this I'll get i'll get back to that in a second uh because it's actually quite an interesting topic um yeah so as you can see I took a screenshot yesterday of the amount of ethereum validators that we have now it's over a million um I question myself what why we need a million validators in in only ethereum um is that really necessary um you know this these other protocols or other systems out there that don't have that amount of validators um especially if you look at kind of the distribution of where these validators uh come from you know a lot of it is from Lio a lot of it is from coinbase binance Kraken rocket pool Etc um meaning that every kind of extra um like stake that you put through one of these entities or through one of these protocols doesn't really add that much to decentralization because you're just adding to one party um and making that party bigger especially you look at that things like Lio for example um or coinbase if you stick through them um so yeah what is the actual problem right so there's actually limits to peer-to-peer networks um I'm not sure if you've ever gone into them but um this there's been a study um by the ethereum foundation that looks at um the amount of validators in the network and when it would potentially to a halt so the um if I recall correctly in at def connect in Istanbul last year they talked about 1.6 million validators you would start to run into limits of peer-to-peer uh networking um and actually having issues with um with like you know sending messages on time and getting around the network as fast as possible because you need to kind of get these messages um like transactions or or even just publishing of blocks uh to other notes as fast as possible um also the benefit the next validator um as I kind of have named it which is kind of what I referred to before is if you add more stake or one more validator to the network how much more security or how much more trust or how much more um decentralization are you actually going to gain that depends on the networks but um depends on which network we're talking about but in ethereum um uh especially there's not that much more uh uh to be gained by one more validator because there's already over a million right so what does one more validator add to that um also especially on custos ecosystems 80% of the stake is be most of most of these networks have 80% of all the stake that is uh States against validators within the top 30 of all those validators um you know that means that those top 30 have a lot of control over the network but also usually those top 30 are trusted companies or trusted entities um that have run nodes on many different networks for Years or or even decades at this point um so this kind of a a question whether or not that's a problem um I personally don't think think it's a problem as long as these entities are um uh trusted entities and you've looked into them or you know the people behind them uh or you've seen what their their infrastructure looks like um High finality time it's kind of referred to with the pp Network limits as well um the moment you start running that many validators um you can run into limits uh you can get issues with finality time right like ethereum has aity time of um 7 or 14 many minutes um while com be based networks have finality times of um seconds usually um there's even examples of noble for example it's a um stable coin issuing chain um on the cosos ecosystem that has you know less than one second finality time and that's like you start you start running into um like salana territory at that point for like speed um which is pretty cool um so what are potential solutions to this right so we've kind of looked at the problems how are we going to fix all these issues um or do we even need to um well ethereum is working on something pretty cool um and they're actually launching this in the next upgrade which is the EIP for uh increasing the max effective balance um I see ended up quite small but um you can kind of get the picture um on the left you can see there's U five different stakers they all have uh one validator which is 32 e um uh running um we there's five nodes on the network and they all have uh one Val data running on them well actually what's happening is the middle part here which is we have one node or we have three stakers but one of those stakers actually runs three validators on one node um so what do having one more that's actually possible this is what a lot of people do um even we do this um so what does having that one extra validator um on top of like what if saer one here adds one extra validator does it add anything not necessarily um because you're just adding one more Val data to the network you're adding some economic security but you know with the amount of Val lers are already there doesn't really add much um so what they're working on is something that's to the right here which is um uh a a model where instead of each Saker having to around multiple validators to take a larger amount of eth and actually gain yield from it um they are working on raising the maxtic balance to actually 248 um if I remember correctly um because they might have changed it recently they keep talking about it in the in the dev um uh talks um but the um 48 would mean that all these validators that are or all these nodes that are running you know 10 plus validators on one node could just consolidate that into one validator um it would mean the amount of peer-to-peer traffic um that you would need to send AKA all the attestations for every single validator um would decrease significantly um so that would in theory help lower the amount of validators on the ethereum network and maintain stability so um something that has been kind of in development for a few years now is um light nodes so light nodes are the way I think about it is um instead of having to use an RPC endpoint from a um hopefully an entity that you trust um so there's RPC endpoints from entities like um uh let's see coinbase or poru or Etc usually those are you know trusted entities but it would be cool if you could actually run run something yourself without it having to take up a lot of resources on your computer or having to sync for a really long time um Etc so um people have been working on a thing called light nodes so what's pretty cool about light nodes is it downloads stores and verifies just the headers of a chain but but just downloading and verifying just those headers um you can actually run your own RPC endpoint while it's still being a very lightweight thing to run like you can run it on a browser people have you can run on your phone um I've run this on my phone it's pretty cool um yeah and it it actually SS pretty fast as well it takes a while of course it depends on the Chain but celesia has implemented this um on their side and actually one of the so what you see here is a screenshot um from an explorer that um is specifically built for CA um that um you can just run a light Lo on on the website and it will just use that light node as its data source so it's really cool you don't need to use their RPC provider um like you have having to trust a centralized entity um I kind of call it web to like RPC providers um so yeah you can you can kind of know trust but verify yourself you don't have to uh uh rely on anyone else so at home staking I think this is a bit of a for me this is a bit of a controversial topic because I in my opinion I don't think that this is going to solve any issue these teams might not like me after this but um the a lot of teams have been working on at home staking So like um SSV nectar obal some others out there that been working on specifically DVT technology stands for um decentralized uh Val dat technology it's a bit of a weird one um so kind of what what decentralized valid technology actually is is it allows you to run in most of these networks allows you to run um an ethereum validator but only with one e State um so what that means is you're collaborating with 32 other at home sakers to run one full ethereum validator um and you could do it at home and the reasoning for being able to do that at home is um if a um validator goes down so let's say your power cuts out um in your house or your internet is out for a day um those those other 31 um nodes that on that are kind of on that uh small like um Network that's running one validator um will make sure that you still keep signing um so users like threshold uh signing you can kind of see it in the picture here but it's a bit technical um most of the time as long as the majority of those nodes are online the validator still runs and you'll still earn a yield um although this system is built in to incentivize you to keep your node running at all times of course um so so my question to this is to what benefit is it to run one e um validators at home right we're we're going to add a lot of stress on peer-to-peer networks by running one by running all these nodes on the network um so running one e uh validators means that you're you're adding 32 nodes onto the ethereum network um just to run one extra validator again you know how much benefit are we uh adding to the N to these networks by running that many nodes with with that that much peri- to period traffic just to add know one validator worth of um uh yeah one validator worth of Economic Security um so yeah I said no extra Economic Security slightly no no extra um no extra Economic Security because you know one Valor doesn't really add that much um in this scenario because we already have a million um and to add to that if you st one e there's actually not that much extra yield like you're not going to earn that much yield most of these people that run is that run these at home are going to do it for fun they're not going to do it to earn a lot of um to earn a lot of money um if you want to do that is actually it's actually more beneficial to S with Lio or with coinbase because they can incentivize you further with different um tokens or other token incentives um other token structures to help you um gain a higher yield cool so another issue with um decentralizing Val says this geographical decentralization so this the thing that I'm very passionate about is um almost one third of all ethereum nodes are located in the US um so you know what happens if if uh actually it's quite worse um because um if you look if you zoom very far into this map into specifically Virginia you're going to notice that this one state in the US and this one town in the US runs a very high majority of all those nodes mainly because of these Cloud providers that are all located in that one geographical area and that's their default kind of cloud location um so Us East one you might have heard of it um so Google Cloud AWS and that's actually hner who's mainly in in Germany and in Finland but actually if you look on this uh list you see Germany as the number two and Finland as the number three so a lot of people are running their their validators on cloud infrastructure there's some danger with that um the moment that these Cloud providers either say hey like we don't really like from a legal perspective or whatever we don't really want to be involved in these blockchains um we're going to start um kind of kicking kicking all the validators out or everyone that runs uh block infrastructure out for whatever reason right it could be um PR drama it could be whatever reason they decide legal issues um for whatever whatever reason if they do decide to kick people out all these validators all these nodes are going to have issues and everyone's going to have to scramble to start running the node somewhere else um which actually that would if you think about that would kind kind of be a cool thing to see happen just to kind of see all this geographical decentralization spread out a bit more um so I actually did a a cool study um not that not that deep but um kind of good enough to get some result some interesting results which is I I looked at our um our validator nodes uh and the pairs that it had um so I I kind of pulled the list of pairs that our validators have or our nodes have um and looked at where the IP addresses that that we have as peers are located uh within Europe specifically um so there's kind of some standouts here um mainly Germany uh probably because of header but also a lot of um uh staking providers and and uh other validator entities lights R the nodes in Germany um there's certain reasons to that um but you know a lot of that's also Cloud um so France is also a big one ovh is located there again cheap Cloud providers um so lot of people are going to be in s are just are just going to take the cheap rout and go into run on those Cloud providers there um so Finland also had some location they run there and the UK um I think this kind of at least me it didn't really surprise that much um isand mainly for SOI but that's just for um I looked at the amount of um I looked at it by Stakes not necessarily per node um so that's it could just be that there's two or three validators there that just have a massive amount of stake um so yeah this I thought this was pretty cool um mainly to see that Germany just very overpopulated with um validators um so actually um I've been working on a project to run our own nodes uh collocated in Bulgaria um some cool reasons for that um barely another not barely any other nodes if you looked at the previous slide Bulgaria was highlighted mainly because there were barely any other fellator or or kind of nodes located in um Bulgaria itself um so they made it kind of a cool pick from decentralization perspective um the internet speeds here are just outstanding it's located in an area where it has amazing connections um to both uh Western Europe where lot of the major internet exchanges are located um and to um uh the Middle East uh and Beyond Asia um so it's just really well located for an internet um uh kind of centralization perspective um there's a high level technical expertise here you know great universities especially in sofhia um means that you know if you ever if we ever have any issues I don't need to physically go to Sophia although you know I live in Bulgaria but it's I'm not geographically very close um so it's uh it's cool to have a very technical uh team that we can always contact to uh solve any issues for us um and it's a good investing climate in my opinion um I definitely see Bulgaria um uh being good investment in the in the long term cool so what are next steps for decentralization so I'm personally um and kind of from a validator perspective working on encouraging other validators to to dive a little look a little harder into their their choices of where to run their nodes right it's easy to pick Germany it's easy to pick Switzerland it's easy to pick the UK or easy to pick the us or wherever else you're looking into um uh running your fail datas or or your nodes for for the different networks um but you know to kind of help decentralize these networks and to prevent any government or any uh inet service provider um from uh causing you any issues um by being able to kick you out or whatever it it does actually make a little of sense look a little further than those countries um so I actually know a lot of Val datas that are now looking at Estonia and other baltics um so yeah there's some stuff happening but even even those countries are geographically actually quite close in Finland for example um so if we want to kind of spread out for where all the users are and where people are that uh want to um you actually use the networks um it makes sense to look at you know where people actually using these networks and um do they um like how much do they use these networks and what's the closest node to them so it might happen that the closest node that someone's running for a certain network is actually located quite far from uh that Center of of usage um it's like cell is a good example they have a lot of usage now in in Africa um recently they've really taken off in the amount of um wallets that are being created there and are active there um but there's not that many nodes located there so like they ran into issues and people are are launching other nodes in a closer to those countries in Africa where um where people are using it um and and I think we should put an emphasis on on light nodes versus home saking home saking is cool to like you know to test your test the waters with staking or like running a node yourself uh running a validator yourself but in the long run um light nodes in my opinion are really going to kind of um help decentralized networks further um kind of for all the I mentioned earlier so there's some other um Concepts in the blockchain space that are working on decentralization um so sequencing for uh L2 networks has been a big thing uh recently so a lot of teams are working on shared sequencing um TDR what that is is um an L2 needs a node to kind of combine or like to collect all the transactions uh create the uh the role up and post that post the the data that it needs to post to an L1 um a lot of these Networks have just run that themselves so like polygon has kind of been one that just R ran nodes themselves to do this they didn't really work together with other um entities but there there's like a centralization risk there right you like one if if their nodes go down what happens and two is um if if they really want to block someone from doing something without that necessarily being uh they're being consensus within the community that's a good thing they could um so yeah shared sequencing is a cool idea to uh help um decentralize these these networks um uh or decentralize l2s a bit further um ZK proving is is a big thing recently Le in my space because I work with it um but um it's pretty cool to see that um these uh uh CK rollups are kind of starting to uh really get into production but there's still an issue with the actual proving uh of those uh um ZK wollups um so um um shared proving like proving networks are kind of helping with um know running Pro running provs and uh do those are actually very uh computationally intensive as well usually um so like we're talking like GPU proving um kind of going back to a little bit of the mining era but um still in a very uh cool and interesting way um also there's recently been a push in in better tooling for high availability for validators which that doesn't necessarily add to decentralization but it does help make sure that all the nodes or all the validators are in the network actually stay online or like stay online or that nodes are incentivized to stay online quite a bit further and longer um so like if if there's an issue on one node you can have a second node that still keeps signing um while there's something on top that helps prevent um double double signing Etc cool so that was it [Applause] [Music] la