0xResearch Cross-Post: The Bandwidth Bottleneck Holding Blockchains Back | DoubleZero

Transcript

A lot of people are like, "Oh, what's the magic?" It's no magic. It's just what did every other tech platform do when they needed to become high performance and scale? They built a purpose-built um custom dedicated network underneath. Quick shout out to Permissionless 4 coming this summer to New York June 24th through June 26th. We'll have speakers from all across the industry, including Hayden Adams, Kyle Sani, Guy Young, and more. Get your tickets today by following the link in today's show notes, and make sure to use code Bell10 to save 10% on your purchase. See you there. Hey everyone, quick disclaimer before we get into today's episode. Nothing said on Bell Curve is a recommendation to buy or sell securities or tokens. This podcast is forformational purposes only and the views expressed by anyone on the show

are solely our opinions, not financial advice. Our guests and I may hold positions in the companies, funds, or projects discussed. Hey everyone, this week's episode is a crossost from ZeroX Research. We sat down with Austin and Matteo to discuss all things. They explain why bandwidth is the real blockchain bottleneck and how D0ero addresses it. Validator economics, token incentives, performance benchmarks, and plans for Salana and beyond. Enjoy. Um, what's up guys? Welcome back to another episode of Xerox Research.

Today I'm joined by Austin and Mateo from0ero. Thanks for coming on guys. Thanks. Great to be here. Thanks for having us. Perfect. Before we get started and actually dive into zero, I kind of want to get a bit of background on both of you and what you guys do over at zero, if that works. Yeah, sure. All right. Uh, so I am one of the three co-founders here at0ero. I am the co-founder that is attached to the foundation side of the project. So, um, I also serve as the president of the foundation here. And prior to joining uh D0ero and co-founding this thing uh I spent the last four years at Salana Labs

and then the Salana Foundation. And I'm Matteo Ward. I'm the co-founder of Malbeck Labs and um Malbec Labs is a core contributor to the zero protocol and network and we primarily do protocol and network engineering. Perfect. Um fantastic. And another uh introductory question, what is exactly zero and how is it beneficial to the general crypto landscape? Go for it. Well, um I guess people on the internet are tired of hearing Austin say it, so we'll try to give it a little different spin this time. Um zeros is a alternate internet, a purpose-built internet for

distributed systems. And what that means is that it's not a general purpose network. And you might find it funny to hear someone describing a project in the negative, but it's very very important to understand the difference between general purpose and purpose-built. And so basically what we do is we say what does a time-sensitive distributed global system need in order to operate at high performance and we build a purposeful network that is exactly focused on delivering that performance. Uh so that's what double0ero is high level.

We'll dive into more of the details of how it works but um you know basically it's really fast a really fast highway for cars to drive on. Perfect. Um and why why do we why do we need it? um in the sense that I read over some of the tweets that Austin put out as well as your white paper and one thing you guys mentioned is that compute is not the uh bottleneck anymore and that um the bottleneck is bandwidth limitations that's the word um what what does that mean um more simply and then maybe you can dive into why zero is necessary I think I'll start with the high level and then I always love to hear Austin

talking about why the bottleneck exists in blockchain specifically. But if you look at the high level, every single tech segment and industry has gone through an evolution where they start out at a hobbyist level work running on whatever internet they can get their hands on. Uh whether it be a land party uh for gamers or whether it be the public internet for networks like YouTube when they started out or Netflix when they started out. And I don't know if you remember if you ever watched Netflix or YouTube pre-private network, but around 2006, Google um deployed their global private network and YouTube started using that and all of a sudden

you stopped getting buffering and delays and bad user experiences. The same thing has happened in cloud networking. um AWS, uh Google Cloud, all the big providers have these huge private networks, purpose-built networks, um based on submarine fiber, terrestrial fiber, all linked together. Uh and so, you know, basically the time has come where blockchain needs to do the same as every other tech industry. So, um a lot of people are like, "Oh, what's the magic?" It's no magic. It's just what did every other tech platform do when they needed to become high performance and scale? they built a purpose-built um custom dedicated network underneath and we got to blockchain and we said, "Oh, blockchain has some specific issues and

that's where Austin can kind of help us get that insight." Perfect. Um Austin, do you have anything you want to add? Yeah, so I mean this is kind of one of those funny things where when people talk about the problems in blockchain today, they're really focused a lot on compute. And one of the questions you should always ask yourself is why, right? Why is compute the thing that gets so much attention and because it's kind of the easiest thing to solve, right? You people solve the problem that they know how to solve and the problem that is immediately in front of their face. And so a good question to ask yourself is why are test net numbers always so much higher than mainet numbers? Right? And there there's two there's two reasons for this. One is

mainet load is different than test net load. That's that's very kind of important to remember. The other though is that test nets are usually small deployments in se like small data centers or they're running in much more controlled environments. And the thing that we've seen from like pretty much I would say mid 2021 onwards are that validator clients now the software that actually runs these things can do very high performance systems. Now if you go back to 2018 and you ran GE locally you weren't getting over you know 200 to 400 transactions per second on the EVM.

That's different nowadays. You look at something like Salana and you can do a million transactions per second on the Fire Dancer client. You can do close to that um on the Agave client as well. And you can actually look at like uh you know how does that actually scale out when you're looking at a test net. And one of the pieces you run into is if you were at Breakpoint last year, you saw a demo of a million transactions per second running on uh you know I think it was six nodes and four continents and that was actually running on the zero network. If you'd run that demo in the public internet you would not have gotten a million transactions per second. And so really we've we've reached this point where for years it was the software that was holding blockchains back. And it's no longer the

software. And so like our our goal at double zero is to like make it the fault of the software people again. Very good. Um one one thing I want to ask about is we typically um say that consensus is an issue coming to consensus is an issue because um let's say New York and Japan need to communicate like validators in New York and validators. Is this what um zero would help out with? that would be much much quicker because it wouldn't get routed through some random. Yeah, there there's there's like a bunch of different ways to think about this and I think probably the easiest to start out with and I'll have Matteo do the harder one is like the pointto-point communications, right? And so if I am

trying to talk to a computer in Singapore from New York, traffic will bounce all around the United States uh you know before it actually heads in the right direction and eventually gets over to Singapore. There's a bunch of reasons for this. The main is business use cases um which is that like the the way that these business models are constructed over the last 30 years is not to incentivize performance. It's to incentivize basically least cost routing and getting data off your network as quickly as possible. And so even just on the onetoone example of I'm just trying to send a packet from New York to Singapore, it's not going to take an efficient route because the business the business models basically don't support that today. Like you'd be surprised to

learn this, but there's actually no way to pay for packet prioritization on the public internet today. Like I can't just say, "Oh, I'm willing to spend a bit more money like a geo tip or something like that and get this thing processed more quickly." Like there just that doesn't exist. Um and then there's a whole other problem which you know Matteo can talk about of the one to many communication style of blockchain which the internet just wasn't built for. Yeah. I mean, just to add on to that, why does the traffic bounce around all over the US, all over the world? Uh, it's related to the way the internet routes traffic. Uh, you can see on my shirt here, this is the uh the current ASN number, autonomous system number that we're using on test net. And it's

one ASN for the whole world. Uh, that sounds simple, but that's a huge statement because on the internet, you've got dozens and hundreds of different ASN numbers. Each one is a different network and they have to pass traffic between them and they use a protocol called border gateway protocol BGP to do that and as Austin said the BGP attributes that route that traffic have many different factors it could even be geopolitical be like don't cross over this country could be use this border router only in Dallas when the traffic needs to go to Seattle it's going to route through Dallas think about it as an airline um hub system right you know if if you're going to

have to take a connection uh on a coast to coast flight, it's never going to be the direct straight line. And so you're always going to go through a hub. It's the exact same thing here. And what the magic of what zero does is it routes the entire world on a single ASN number, which means that every single packet goes straight to uh the location that it's going to. And there's peering between every node on the zero network and every other node on the zero network. And that is not the case in the internet. There's peering at the edge. There's peering at these specific points. Carriers fight with each other and then they depear and then you can't ever deliver traffic directly to that city. There's so many different things that happen. And we've solved that by

having this single mesh of peered routers all around the world under a single ASN number. And we've done it in a decentralized fashion with multiple network providers, which is the key because any other network that's like this one is a centralized privately owned network. And we've brought that into the public domain. We've made it decentralized and we've made a whole community of contributors that make up that unified network. Yeah. And this really is like where you need tokens, right? Like a lot like we get asked every once in a while like, oh, like do you really need tokens? Couldn't you just build this as like an equity company? The answer is like no because we need the ability for everyone in the link of the chain, for lack of a better

term, every single network provider that you pass through to get from New York to Singapore in the most efficient route possible to get paid. And that that's like sounds like such a basic insight, but like that is not how the internet works today for the most part. Um, and so being able to have these things whereas Matteo was talking about the incentive of every provider is to provide as most useful service as possible because they'll earn the most rewards, right? The links that are give the biggest advantage over the public internet are and reach the most amount of stake or validators or RPCs are the ones that are rewarded most on the zero network. And so the incentive for everyone is how can I make my links as performant as possible and as useful as

possible, which like again sounds like the way the internet should work, but it's not actually the way the internet works. And this is really only possible with tokens. Today's episode is brought to you by Ledger, the global leader in digital asset security. Celebrating 10 years of innovation, Ledger is trusted by 7 million users, secures 20% of the world's digital assets, and remains the most secure hardware wallet for your crypto and login. Follow the link in this episode's show notes to grab your ledger today. I'm glad you guys brought up both tokens and contributors cuz I want to ask about this. Um, Mateo, maybe first to you because you brought up the contributor side. What who are the different stakeholders for zero? Who are the involved parties that you guys have to deal with and that you guys have to

coordinate with? Yeah, let's start up high level. It's anybody that has network that's high performance, low latency, low jitter and wants to contribute it to a global new version of the internet. And when you look at that high level, you're like, wow, this is a huge market opportunity. Um, when Starlink was launching their project, they they took the TAM to be a trillion dollars, which is the annual amount of revenue on the internet. A trillion dollars. They wanted to get 3 to 5% of that 30 to 50 billion and today they're at 11 billion. They're doing fine. Um and so there is a massive global market for connectivity and

there's lots of people that own capacity. And so a new use case for applying that to um a monetizable project is wonderful for all of these uh people that own networks. And so really the the addressable contributor base is anyone who owns and operates a high performance network that wants to be part of a new internet. Um zooming in and making that into a timeline, you get different phases. You get the early adopters and those early adopters are the ones who probably understand blockchain. They understand what a utility token is and and the value that might come out of that. Um they have underutilized capacity. So they've got a global network that's being used for

another use case in house and they have a slice of that that they could contribute. Um and they're risktakers and and people that want to believe in something new sort of bleeding edge uh tech adopters. And this is a really fun group. These are the the ones that have initially committed to to the network. You've seen them in some of our press releases. You've got uh companies like Jump Crypto. You've got um uh other VCs that have uh contributed. We've got Rockaway X. We've got um Distributed Global. We've got a number of people that said, "I'm going to make a bet on this because I think it's the future."

And then you've got phase two, which are the wait and see ones. We're already starting to get some of those where they say, "Oh, wow. I see this actually works. Um there there is value here. I can create a business case. I can do uh an ROI analysis and I'm going to go in that and that includes also carriers and other enterprises. So there's uh large enterprises that may not be high frequency trading firms or others that just have a network and want to contribute it. Um retail companies and others, they all have private customuilt networks because they all need to move data around very quickly. Um, and then you've got phase three, which is equally interesting, which is entrepreneurs, capitalists, and all kinds of people.

They're like, "Wow, I've got this ecosystem now that's a a tokenized version of telecom t capex and of shared networks, and I can pull together the pieces myself. I can go out and find someone who's going to provide me with data center space. I can find someone who's going to lease hardware. I can find someone who's going to lease me connectivity. I can pull it all together and contribute it to the double zero network and earn tokens on that and and that's really the the the end vision that that this really becomes viral because um after the institutionals if you get this thriving ecosystem of entrepreneurs um contributing assets to the network it really scales quickly. That makes sense. So the first I guess users that will be onboarded

apart from those who have already committed are mainly people who already have these resources set up and that are just willing to give instead of people this isn't something that I could do for example until later on when I can tell that it's going to be very valuable. Yeah. I mean this is you can think about this as like a a professionalization of the role of infrastructure in blockchain, right? So you start off with like a Bitcoin miner that you can run on a Raspberry Pi and you can run on a generic computer and nowadays like you really can't mine Bitcoin unless you're investing in AS6 and you're in an area with cheap energy and you're putting significant you know capex and opex into that operation. You had you know Ethereum where like sure in theory you

can validate on a Raspberry Pi but you can't really like if you want a full node it does still take a a reasonable real computer. Um, but that network is quite slow. And you look at anything else, whether it's Suie or Aptos or Salana or any of these like L2 sequencers out there, they're very heavy machines. They're basically professional operations at this point. Um, and so individuals and small companies can certainly run uh infrastructure for all of these networks, including double zero, but it's not really the type of thing that you're expecting like an average home user to do because quite frankly, these economies have just gotten so profitable and so performant that this is like a mark of success that like this is not a hobby for anyone

anymore. This is like a full-time job to be an infrastructure provider for a modern blockchain. And so double zero sort of asks the same out of folks. We we do have you know small small firms that are providing capacity to double zero. We have uh you know two firms that have never done this before right X never had any private fiber. Distributed global never had any private fiber as well and they're they're contributing uh you know to the zero network at this point. So, it's not like you have to be like a jump trading or a galaxy or a terra switch to be able to contribute capacity, but you do need to be putting in real resources to this because this is mission critical infrastructure for networks that build on top of it. Yeah, as it should be. I'm I'm not a huge fan

of the the local stakers, which I assume will find um support here. I um Austin, you mentioned um I'm happy you brought up tokconomics because that means I can ask about tokconomics. Um we don't have to get into too much detail. You can always say you don't want to talk or it's not decided yet. Very acceptable. Um that's true. I saw that. um what typically I guess deepen um style projects where there's incentives for network operators they face an issue where if you pay out in the native token sometimes it can go down in value as we see um within all currencies and that can be an issue for I guess operators is

this something that you guys thought of where you wanted to maybe go with USDC instead am I doing something that's not the problem with deepen the problem with deepen is not that they pay out in the local currency The problem with deepen is there's usually no demand side. There's only supply side, right? And any market where all you have is supply side, you're going to get a downward-looking graph, right? Um and so, you know, one of the things that I think is really important with with D0ero, there's kind of a few different things we've done. For example, there's no base level contribution or reward.

Just for running a piece of fiber, you by default get nothing. If that piece of fiber is not useful, if it's not providing connectivity into areas where people run uh you know Salana validators or RPCs or any other type of infrastructure on the network, you're not going to get rewarded just for passively having fiber. Most DPIN networks, they give you sort of this passive coverage or capacity subsidy um which encourages farming. And it's um it is extremely hard to farm double zero tokens without providing actual value to the network which at that point is not called farming. is just called being a good network contributor, right? And so a lot of this is just like good economic and system design. A lot of this credit goes to Nihar who is our chief economist

over at the foundation and has done a ton of work on this type of thing. Um but the other piece too is like we are charging validators to use 0ero. We're charging RPCs to use 0ero and so the network once it launches in mainet assuming people adopt it will have revenue inflows to it. And so, you know, the the way that this works is it's not like we're just minting a bunch of tokens every month to give out to network contributors, right? This isn't an this isn't a network that operates that way. Um, validators and RPCs pay to use the zero network. Uh, those are rewards are, you know, if they're paid in native soul or if they're paid in zero, either way, they end up in this reward pool. They're all converted into zero tokens. a portion of those are

burned and then a portion of those are distributed out to the network contributors. So, it's not like we have, you know, like a 10% inflation rate and we're just like printing a bunch of new tokens for network contributors. It is revenue derived. Um, and we think this is like a good solid framework to build this stuff out. Um, we do have a property right now where a number of our uh contributors are also investors in the project. And so, you know, there is obviously with any deepen network, there is a cold start problem. And that's sort of the way we're we're working on a bunch of that cold start problem. But, you know, these are these are the type of systems where like if we're not doing our job and the network's not generating revenue, like that's on us and that means we haven't done this thing

correctly yet. And so I think the um you see this very web two model in a lot of like networks nowadays of grow the thing and like uh you know we'll figure out how to make it valuable in the future and that works if it's useful right like I think like the thesis for Salana and like this was always to thing when I would talk to him about this back in like 2020 and 21 was like or end of end of 20 and mostly 21 was like we just need to make Salana useful and then we'll figure out how to make it valuable and The problem with most DPN networks is they're not useful. They're theoretically useful, but their practical use is very low. You can say this about not just DPN networks, but a lot of L2s fall into the same category.

A lot of new L1's fall into this category. As long as useful and so that means we have validators and RPCs choosing to connect to it. We have networks building on top of it. Um, and we're able to provide, you know, increased bandwidth and reduced latency. the monetization side of things. Like I hate to say we'll figure it out because there's actually a lot of like stuff in the white paper and on the econ and economics.0.xyz about like how these things really will work. But like in some ways like Salana base fees on Salana were extremely low and they still are extremely low. Prioritization fees have addressed that. Bunch of MEV fees have addressed that. There's lots of different systems once you have a useful network where you can you can figure out

what monetization of that really looks like. Our focus is building something that is going to accelerate crypto by 10 years. Um, and if you can run Salana at 50,000 transactions per second by the end of the year and then you can run it at 500,000 transactions per second by the end of next year like uh you know at that point the economics of the network kind of solve themselves. Ledger is the global leader in digital asset security, trusted by 7 million users and securing 20% of the world's digital assets. In recent years, Ledger has built a comprehensive ecosystem of products and services, making digital ownership more secure and accessible for everyone.

Ledger's newest devices, Ledger Stacks and Ledger Flex, introduced the world's first secure touchcreens, making it easier and safer to manage your assets. Alongside Ledger Stacks and Ledger Flex, they have also launched the Ledger Security Key app, offering a safer alternative to traditional password management. Follow the link in today's show notes to get started with Ledger today. One thing I want to ask about that you mentioned is that validators uh and RPCs and whoever uses the network will pay you. Um, which makes sense. Um, I remember there being a proposal about um the double zero take rate and it was 5%. Um I rem another thing is jetto I guess started off they also had a

coldstar problem like you said most of these networks have coldstar problems um and they had very low low take rate I think they ran at a loss for the first until they were adopted is that something you guys are also going to do I assume the 5% is not the initial rate is that correct you know it's a it's a it's a phase in and we have to sort of earn that from you know the network I would say like if if uh if zero makes Let's just get let's just fast forward to a world where D0ero makes validators twice as profitable. Um I don't think there's going to be a hard time asking them for 5%. In in a world where Double0ero has not shown itself being able to make these validators like twice as profitable yet. Sure, maybe that's a big

that's a big ask, right? Um but part of this too is I I like from a deep philosophical standpoint, if you work at a protocol and you're listening to this podcast, turn the fee switch on. Everything before the fee switch is just laring product market fit, right? It's so easy to get people to do stuff on your network when they're not actually having to pay to use it. And so it's the same it to me it feels very spiritually similar to like lowflat high FTV tokens or tokens that launch before a network launches. Like it's it's not really serious until you're asking people to pay for it. if you're just giving away the service, like you haven't found PMF.

You've found if you open up an ATM on the side of a street corner, you're going to get a big line. Um, and so this is part of why I I feel very strongly that we we are going to start with some fee on the network. It's not going to be, you know, 5% of inflation. Um, but, you know, the 5% of block rewards seems like a very good place to start for this type of thing. Um because at the end of the day like we are providing a service uh above and beyond what the public internet can provide and that is we believe worth paying for and if other people don't find it worth paying for that doesn't mean it's not worth paying for right that means that we haven't made the use case correctly or the utility is not quite there um but

fundamentally this cannot exist as a free network that physical infrastructure is expensive this is something that I I feel really strongly about and I've been talking to some friendly validators and debating it and they ask well what about the fees and they say I don't know you tell me is isn't this your network this is not my network that we built this network for blockchain this is a blockchain native network and this is the most efficient way to mine bandwidth or extract bandwidth out of these networks that I have ever seen in 30 years of networking. And so basically what we're saying is we're going to help

you with this protocol extract bandwidth out of these networks at the most efficient way possible with marginal cost with no big capex stepups like every other person who builds one of these networks needs to pay. And um we're going to charge a nominal fee that's going to be funneled back to these um network contributors that are um working together to provide you this shared infrastructure. And it's going to be a super efficient model that we're all going to participate in and that you're going to have a voice in. And this is the wonder of blockchain. We're all going to be moving this forward together. So basically, it's like it's all up to us, the validators, the network contributors, the people working on the protocol to make this economic model the most efficient possible. And

if we can't do it, then then nobody can, you know, because this is a a really great group of of validators we're working with. The network contributors are brilliant. They've built some of the best and fastest networks in the world and so I think we're really well positioned to have a great economic model as we move it forward. Perfect. Um, what does the performance improvement look like compared to like if I'm if I'm a Salana validator and I'm using double zero versus I'm a salon validator who's not using double zero, what what am I saving on costs? Am I is is the TPS much higher? What does that look like exactly?

There's a technical way to to understand this which I'd like to go into first and then maybe Austin can talk a little bit about what we expect um in the pro at the protocol level. But um first of all to answer your question we expect a significant improvement in agg improvement in aggregate performance. Significant has multiple metrics. There's how much uh CPU uh processor time do you free up? There's um how much how many more transactions can you process in a block etc etc. But what what I want to go to is how does this system work from the ground up um from an architecture from a tech stack perspective you take the hardware that

the fastest and best and most efficient networks in the world used today. They're um very uh advanced low latency switches. Uh and and then you put you connect them together with fiber on the fastest routes that are available with low jitter because there's crappy fiber routes as well that have high jitter, but we focus on the high quality fiber routes that are uh wavelengthbased preferably. And so you've got this super ultraast pointto-oint mesh all around the world. And then on top of that, we build the filtration stack with FPGAAS.

Um, we're implementing AMD V80 FPGA cards that are basically a bump in the wire and that can efficiently remove traffic that is not meant for processing, right? It's basically like a green light, red light bump in the wire. Traffic goes over it. If the packet's not signed properly, if the packets a duplicate, it it gets discarded and the and the proper packets um are forwarded on. And so that's another area where there's a huge efficiency that's going to be created for the validators because they're not going to have to process that traffic. And then we we go to how we route that traffic which again is way more efficient. As we said we are using um a full mesh pointto-oint fully peered

global backbone that's going to go very efficiently. So when you combine that all together, we know that there's an exponential increase in aggregate performance. Uh we know that not because of our testing. We know that because every other industry that has implemented this has experienced an exponential increase in performance and we've worked in those industries. So bringing that here, we know what's going to happen. What we don't know yet is the exact percentages. We can guess. We think there's going to be probably 7 70% um reduction in traffic that has to be processed. The garbage traffic has to be processed by validators, but we're not sure. We have to wait and see. We expect there to be 30 to 200% latency

reductions at an aggregate level. The reason the range is so big because is because the internet has all kinds of different ranges of uh of latency and huge amounts of jitter. Uh and so when you put that all together, we're extremely confident of what it's going to be. And um we invite the whole community to watch publicly as we go through the test net into mainet and publish these statistics because these are statistics that we're going to publicly expose. Uh we're already seeing some of that. Um we can probably talk about it later, but today is a new day for0. We just finished a test net upgrade um to what we call IBRL mode. So validators are reconnecting to the network in the new in the new version of

it using a single IP address and we're getting a lot of uh stats and graphics and and traffic analysis that we're going to be able to process and publish over the coming weeks and months to show that aggregate improvement. Perfect. Um before we talk about it on the protocol side, I want to understand what um garbage is in that sense. The the things that you weed out with AMDs. Um what does that look like that doesn't go to the validator? Because in my mind I it's not signed correctly. What does that what does that exactly mean? Yeah, totally. So you got to think remember this is all public key, private key cryptography. That is what all of these blockchains are built on. So, uh, if the

signatures don't match, right, if there's a mismatch, it fails signature verification. And so, stuff that fails signature verification, uh, are are bad transactions, right? So, like if I sort of pretend I'm signing for you on your wallet, I can't unless I have your private key. But a lot of people send those transactions through, right? And it's usually not because they're actually hoping that they gets included in the block. it's that it actually gums up the works on validators, right? And so you can think about this from the perspective of like there's there's a few ways to make sure your transaction lands. You pay a high fee, right? You like use jetto or something like that or you just make it so no one else can send transactions. And when we go back to

periods of time where Salana was suffering from congestion or the Shanghai attacks that Ethereum suffered back in 2019, was that I can't remember exactly. Um those are those are examples of basically economic denial of services. Um and so if something fails signature verification, it is ineligible to be included in the block. It it simply has no chance of ever or it should have no chance of ever making it in because it'll be rejected by the rest of the network. And so we can do things like perform signature verification in hardware on FPGAAS at 100 gigabits per second. And what that basically frees up is a ton of a validator's resources to not have to worry about doing signature

verification on everything that comes in. Now, they're going to do redundant signature verification on what they choose to include most likely, but they stop having to sort through this giant pile of invalid signatures to just find the ones that are valid signatures they would want to include. The other thing that we can also do is uh packet dduplification. And so, you know, sometimes what happens is someone sends a 100 million copies of the exact same packet to the Salana network. Um, again to try and overwhelm the resources of a validator as it sorts through these things and says, "Oh, I've seen this before. I don't need this. Oh, I've seen this before. I don't need this." That takes a lot of time, right? Again, we can do this in FPGA at 100 gigabits per second. and you know doing it there. It

basically means that what validators are getting at the end of the day is a clean stream of transactions doesn't mean all the transactions are high value. It doesn't mean they're all quote unquote good transactions. They could be a bunch of stuff that actually will fail an execution because remember a failed transaction on Salana doesn't mean the transaction didn't execute. It just executed with no state change. So for example, I try and buy soul at a price that is the market has now moved from. Well, that transaction it's well it's it's formed correctly, but it's going to fail because no one will sell me soul at $20, right? For example. Um, and so this is uh basically the way that like0ero can help make these systems more

performant is we can act as this filtration layer to throw out all of the stuff that is no chance of ever making it into a block to begin with. Now we look forward into the future and like there might be other things we can do in FPGAAS as well. What if we are now able to say, okay, well, what if there's a CU to fee ratio above a certain amount? Well, we're going to flag those as potentially high value transactions. And then a validator, again, validators are dictators, right? They have full domain over whatever they're building uh for that block has to be voted on by the rest of the network, but they could include whatever types of transactions they want or build even an empty block if they want to. Um, and so I we can over the long term start providing a

bunch of additional value ad data to validators about the in the incoming stream of transactions that they're getting to both filter out garbage and to help them make better decisions about profitability. That makes sense. Would you um would would it be possible to add a um bidding mechanism to something like that where if there is high valuable transactions that you can ask people to bid for them or is that something that doesn't make sense? If FPJs don't really operate statefully in that way. Um so these these are sort of like you you can think about this as like there's a ton of stuff a computer does that's like sticking something aside to look at it later, right? And this is like an

auction system. like, oh, we're going to we have a slot here. We're going to take bids or whatever. Like these these operate more almost like filters. You can sort of think about them that you can do a ton with an FPGA, but the the way that we have these things built is it's much more of like a pass don't pass or pass with some modifications attached to it because the goal of the FPGAAS and the reason we're using FPGAAS is they're incredibly fast from a latency perspective for the type of operations we're trying to perform. So this doesn't add like if you're adding an auction in line suddenly an auction takes time to run so that adds latency and our whole thing is increase bandwidth reduce latency and so someone could like what we would basically say if someone wants

to do that is that's jetto or you should build a goto competitor on top of the zero network but you're going to need x86 machines to do all the normal things you'd want you know compute done on a network for it's not really something that you'd handle through the FPGAAS. That makes sense. And one thing you mentioned is that you guys process that at a 100 gigabytes per second. Can you put that in comparison to what a val How much space are is this freeing up for validators? Yeah. So most validators today are operating with somewhere between a one and a 10 Gbit internet connection. Um and so this is anywhere between 10 to 100 times the capacity that they

currently do. The largest DOS attack that Salana has ever sort of suffered that we have metrics on was 80 gigabits per second coming into one data center. Um, so this, you know, I think this is this is uh more than we've ever seen the Salana network be be attacked for previously. Um, which is pretty awesome. Um, Matteo, do you remember how many actually transactions per second 100 gigabits is that we can sig verify? No, I'm not sure in in the Salana protocol, but yeah, what I wanted to add to that is that with 10 to 20 lines of code, you can define what you want to do with a packet in this wonderful software that that has been built. Um, Jump Crypto did a lot of the the development on that.

And um so you can go in there and say very simple code. I want you to do this with the packet. Uh and you can adapt it to other protocols um other blockchains, other use cases, whatever. It's sitting there as a filtration layer in the network and it's extremely powerful but um there's definitely a huge capacity to to do transactions in there and um they they'll scale. These are not going to stay at 100 um technology scales and and they'll scale up and we'll be uh able to compete very effectively with with centralized systems like Visa or credit card processing or um uh NASDAQ or anything that processes millions of transactions. This episode is sponsored

by Catria. Catria is the first zero knowledge rollup to enhance the capabilities of Bitcoin block space, enabling Bitcoin applications, and unlocking Bitcoin's full potential as a programmable financial platform. With Catria, developers can now tap into Bitcoin's liquidity to build scalable payment systems, privacy solutions, and more. Citria is currently in its test net phase and is approaching its mainet launch. To learn more about Citria, follow the link in today's show notes. One thing you mentioned is that it can be customized for other chains. Um you guys in your white paper talk about being able to use this for other L1's, RPCs, other L2s.

Um what does that look like? Um I guess or do you have anything on the road map that you're like we want to focus on this? Um is this for any chain or you going to mainly focus on Salana for the next year, year and a half? I mean Salana is the place where there's the highest product market fit at the moment because it is built in such a way that it needs a it like the validator client is high capacity the server like it's a lot of validators right it's the largest validator set for a high performance network in the world um with about 1,300 validators at the moment and then probably another 700 RPCs so 2,000 different nodes around the world there's actually like a ton of use cases there

and and like the more nodes you have the more you need d0ero and we haven't even talked about multiccast yet that we'll get into that in a minute that's a whole other use uh for this um but Salana also has a ton of activity and is very competitive activity and so networks where um there's no competition the filtration services aren't necessarily that useful right like and like this doesn't that doesn't mean that they're worse networks but like Celestia it's a DA layer there's not a lot of like competition to write to the DA layer, right? That's just like it's not it's built for a very different type of use case or Filecoin and IPFS. like these are just systems that don't have competition in the same way that like Salana does, you know, and so this is

part of why we have this IBRL mode which we actually just launched uh today on testn net which is a mode that actually doesn't have the filtration services in it in the same way and that's not designed for I mean you know in testing we're testing it across everything but that's designed for networks that are both globally distributed they're high performance but they don't necessarily have this competitive aspect to it. You know, right now there's huge amounts of competition for a Salana block. There's not huge amounts of competition for a DA layer or a CDN or a storage protocol.

And so, um, you know, we're starting with the hardest thing first, right? That's like that's very much in our DNA, which is Salana. And so if we can nail the Salana use case, um, expanding out to other networks and customizing FPGA code to to meet whatever they need or expanding out to these non-competitive use cases that still can benefit from IBRL. Um, you know, if we nail Salana first, the rest is downhill from there. And I assume competitive use cases here means there's a lot of money to be potentially made if you get into a block and that's what you guys are going to be like chains where that is the case or I guess me relays maybe RPC. Yeah. So like uh if you don't have competitive markets

you don't have capitalism. You've got something else, right? But like like at the end of the day, the reason that trading on a centralized venue is great is competition, right? The reason that companies like Jump and Citadel and DRW have dedicated fiber lines between New York and Chicago and all sorts of other technologies they don't even tell us about um is because there's competition and there's monetary premium and advantage in being first, right? And if you don't have those dynamics on a network, that means just there's not much money being made on a network, right? A network that is not competitive is a network that is largely not profitable. Um, and so that's one of the best things we can look for is

competitive trading systems where like I not only have an incentive to try and beat you to the trade, but I have a reward if I beat you to the trade, right? I make more money. Like you look at like high frequency trading and it's like the fastest person makes 90% of the value and then the second fastest person makes 5% of the value and the third fastest person makes two and a half%. It's it's an extremely sharp declining curve in terms of like value relative to speed. And so that's one of the characteristics of Salana too. Um and any network that has that same type of characteristic is one that rewards performance. And networks that reward performance are where D0ero has highest product market fit. Not to say we're not

going to support all other types of use cases as well, but like you know, you got to chase the north star first. Yeah. No, that makes sense. I feel like that speaks to the general shift in crypto that we've had as well where increasingly it's become more and more competitive like we have chains now with 20 validators just because it allows it to run a lot quicker and similarly we're seeing more and more HFD and collocation. Yeah. And so I think this is like this is where it gets into it I think it makes sense to talk about multiccast for a minute because the reason these networks are going to 100 validators or 20 validators. It's not like it's not because they're evil and they want to only reward 20 people for running validators. It's that there are

like real exponential data problems the more validators you add into a system. And so the faster you try and get, the fewer validators you can get away with, the easier that thing is. And so this is part of like the whole premise of double zero is rejecting that premise and proposing an entirely alternative way to move data around that doesn't require validator stats to get very small to also achieve high performance. And that's that's multiccast. And you know, uh I think we should have Matteo talk a little bit about that.

Yeah, just it's a traffic propagation problem. As simple as that, right? The more validators you add in more places around the world, the harder it is to propagate the traffic out to them. And blockchains, because of the the culture of decentralization, they want to be distributed. And so they the more distributed they get, the slower they get. It's it's just a tragic reality of uh decentralization. And so the way that you fix it, surprise surprise, is the same way that the stock market fixed it, right, is using multiccast. If if you're um collocated at NASDAQ, you're going to get a multiccast feed um with market data on it because that's the fastest

way to propagate data out. Now, what's the problem with multiccast? Multiccast needs to run on a single ASN. It cannot run on the public internet. And so that's again that's the magic. That's what makes us different from a lot of deepin projects that have pieces of infrastructure that just get um contributed and then used. All of the pieces are part of a whole in this case. They're all making up this single ASN and that allows us to run multiccast over it and that's um very very efficient compared to the way that um normal traffic routes in blockchain.

I've been reading some white papers. I love to read new L1 white papers when they come out. They're like, "This is the fastest L1 and and I'm like, oh, well, how do you propagate traffic?" And then I read the traffic propagation piece and it is this crazy compression thing trying to fit everything into these tiny box limitations of available bandwidth and throughput and then they infer statistical conclusions about what might have been propagated. No, just propagate it the way that traders propagate traffic on multiccast um with, you know, a lot of bandwidth, low latency, and using the multiccast groups to get the traffic out to everyone who's subscribed. So, we're doing this um we're right now in in testing mode on

multiccast uh with JTO and um it's going well. It works. um we can we can multiccast shreds out um very efficiently and um it's it's revolutionary from a networking perspective in blockchain because as far as we know this has never been done before and it truly is uh the fastest way to get traffic out. I'm glad I'm glad we got two references to the t-shirt and I'm glad that you guys pointed out that it just makes sense to do what Trafi has been doing. I feel like every year in crypto we figure out like sometimes they have come up with good things that we should just copy and make it more decentralized and make it work. Um I wanted to ask about this deepen label

because I want to get do you guys like being referred to as a deepen network or is there something else because I know you call yourself N1 as well and I always I guess when I think of zero I don't think of a deepen. Um but is that what you guys are going as? Yeah, I mean this is kind of tricky because I think I think uh deepen usually is a system where it is lower quality service but higher reach of service. If you look at Helium, like it's not as powerful as a cell tower, but you can put it everywhere, right? Hivemapper, they're not as good cameras as Google Street View, but the frequency of data is so much higher, right? Most deepin networks have that characteristic. Double0ero

actually has the opposite characteristic. It is much higher quality than the public internet, but it doesn't go everywhere the public internet goes. Right? The public internet, I can get it in my house. I can get it in Antarctica. I can get it on the top of a mountain. you can't get double zero in these places. Um at le maybe maybe Antarctica someday. We'll see about that one. Um but you know it's not going to be a residential service that comes into into your house. And so I think DPIN is probably the best category name we have at the moment but it's not a uh direct fit. Yeah. Yeah, and that's why we came up with the N1 idea because you know basically it's a base layer of infrastructure and um for us that's sort of the N1 idea network

one level one layer one um we have a really cool graph of the OSI model that shows like layers one through seven and shows blockchain up in four through seven and and double zero and one through three right we're like a low-level infrastructure network and so it's it's sort of a new category but um I would say that it's certainly related to physical infrastructure and it's certainly related to multiple contributors. The difference is the consumption I think as well is that it's like a single um it's a single product that gets consumed as a high quality product uh and it's a sum of the the individual parts.

That makes sense. Um, will there be N2s or will you in Salana fashion call it N extensions or something along the lines? No, definitely N2s. There are several ones already in development. Um, and here's here's Matteo's OSI chart we were talking about. Yeah, thank you. So, you can see the N2s there. Um and uh those those N2s are all kinds of different applications. There's so many different pieces that are needed to make blockchain performant. Um there are things like file storage. There are things like coordination of AI um training, coordination of inference

models. There's stuff like um precision time, geo awareness, all kinds of different things that are needed for the evolution of blockchain. Multiple concurrent leaders requires a lot of this stuff. Multiple concurrent leaders requires some precision timestamps. Uh they require some geo awareness. So all of these different things need to be built on top of uh the base layer of infrastructure. And not only are we going to help um those people that want to build N2s use the double zero network, but we're going to try to help identify opportunities for those things and and bring builders in to to build them. So it's it's exciting.

I are we visualize a whole ecosystem of uh N2s being built on top of of D0ero. Perfect. I have two questions um before we wrap it up. The first one is kind of introspective and I want an honest answer. Is there anything that you guys would have done differently um when you guys set off to start double zero? Is there anything that has been a huge bump that you were like we should have just completely avoided this and instead gone a completely different way? I don't nothing really major comes to mind. I think I think one of the things coming from the Salana world that I underestimated or maybe overestimated is the right way to put it.

Um I thought there was a lot more activity happening on other chains. Okay. Honestly, like I I I know that sounds very like Salana self-s served, but like like our our thesis was that every high performance network needs these filtration services and that that was going to be a major major benefit to all of these networks out there. Today, there's actually very few networks that have the volume of throughput and inbound where they actually need that filtration, which is part of why the IBRL mode was something we decided to build out more quickly than the original plan for it. Um because it is just it's going to take a little while for those and I I think they'll get there, right?

This is not like a oh no one else can compete with salon like everyone's going to be able to compete with Salana at some point. That's the beauty of capitalism um is when an idea is good enough a lot of other people start to do it as well. Um but I think it might take 18 months or something like that for that level of activity to really get to the point where those services are truly required for those other networks as well. And and from the network side, I would say I would have wanted to go directly to a 100 gigs backbone instead of 10 gig. Um we started out with 10 gig, we deployed a lot of um Arista boxes all around the world at the 10 gig level and when we got out into the market talking

to people, people are contributing 100 gig global networks, which is huge. It's we we maybe didn't imagine that people would see it so fast. Um and so we've now got multiple 100 gig backbones being built um all around the world and those require a different type of hardware that's being deployed and so we had to do that go through that upgrade. So if I did it again I would go directly to 100 gig. I'd be even more bullish on it. Interesting. Um, Austin, to your point, um, I think a lot of chains have just decided to instead limit the validator set, and I think we'll see maybe Monad and a few more that are like trying to be both software optimized and larger networks. Um, last question, what should

people expect over the next 6 to 12 months, maybe 18 months? Um, and uh, where can people where where do you want people to go to when they want to learn more about 0ero apart from this podcast? Yeah, so.xyz uh has links to become a bandwidth contributor. Uh, you can sign up to run a validator like a Salana validator or an RPC um on testnet. Uh, you know, six months is hard, right? Like in some ways like six months like I hope we're in mainet, right?

Um, on the 18-month timeline, I think if we are not providing capacity for networks to double and triple and quadruple the available resources on those networks, then I think we've either done something wrong or the ambition of the space is just too small. Right? I think the thing that we've seen with Salana and a few other networks that are very hungry is like and you see this coming back to the Ethereum L1. It's really actually cool to see folks on Ethereum stop paring this idea that the L1 is not for users and start actually focusing on building something like truly useful um for for folks. Like that gets me excited that we are about to enter like a high

performance blockchain renaissance and we're going to see people taking performance very seriously. we're going to stop with kind of like the BS blog posts about how many TPS we can do and it's going to be we're going to start seeing like a real intense market competition on this stuff. Um, and I think that's going to be awesome. I think also we'll see more specialization and like double zero like allows for much more specialization because suddenly the network stack is not commoditized. the network stack is like very high performance and it means that like right now for example if you're running a blockchain that does 50 millisecond block times networking is 90% of your time, right? And it's also very hard to

actually make something that's like very fast profitable when you have so much variance around how long it takes data to get into the system and responses back to the system. This is why like if you notice like HFT firms, they don't really invest much in like the physical computer that sits at the exchange. They invest a lot of money in how to get the data to that computer so it can make the right trading decisions. But it's not like they're all building like quantum chips and putting them in data centers. They're mostly focused on how do we break the speed of light, right? How do we move data as quickly as possible into these systems? And so I think if we're able to provide those resources to software developers, people are kind of come up with entirely new architectures

for blockchain that maybe are even fully hardware accelerated and are just data in, data out, and they're able to use 100 gigabit networks of double zero and multiccast and determined latency and all this stuff to build blockchains that are orders of magnitude faster than people think is possible today. Um, and that's kind of the main thing for me is just like I think I think this industry's gotten a little bit complacent with performance and it needs a little bit of a kick in the ass. Mate, what about you? What are you looking forward to? Well, I'm just looking forward to adoption and traffic on the network. So, I've got like a visual model that I can share with you uh about that. If if you think about today, the zero network is a big ring. Um, and all

around the edges you've got internet connectivity. And what's happening today is that you've got these hair pins happening at the edge. Traffic comes in, it finds out that it needs to go to a location on the internet. So, it hair pins back out and goes to the internet and then hair pins back in. Think about it like a flower, like a daisy with the yellow part in the middle and then all these petals all around the edge, which are the internet hair pin. And as more validators join the network, those pedals start to fall off the daisy and the hairpin stops and you start to get the native transit, including the multiccast uh distribution of traffic super efficient in the core of the network. Eventually, you've only got that middle piece and we've got all of

the massive value unlock with the aggregate performance increase that I talked about earlier. So, that's really what I'm looking forward to. And if you break it into sort of quarters, um by Q3 we expect to have a lot of validators running on the network, a significant amount of stake on the network, and we're going to launch mainet beta as well. Uh and and we'll be sort of in prime time. And then by the end of Q4, we expect to have the the real aggregate performance bumps that you see with all of those validators running natively on the network and building blocks natively on the network. So, you know, I'm really looking forward to the end of the year.

I'm really looking forward to Breakpoint going uh showing the the data, the stats that that we are talking about that we're collecting and having that huge performance bump that, you know, we're all working towards. Very good guys. Um, thank you so much for coming on. This was fascinating and something that I did not know enough about and that now I actually think I can say that I understand at a good level. So, I really appreciate you coming on and taking the time to explain and talk through. Wonderful. Yeah, thank you for having us.

[Music]