Building a New Internet for High-Performance Blockchains | Austin Federa - DoubleZero

So, I'm gonna get that down, right? So, when to walk out on stage. Hello, everybody. Everybody awake? Can I get a whoop whoop?

All right. You guys are beautiful. I love seeing you guys. Okay. I want you guys to welcome Austin Federa to the stage like with a real shout, applause, and a whoop whoop.

One, two, three. Okay, Austin Federa.

There we go. That's a bare market whoop whoop if I've ever heard one. Um, well, hey, I'm I'm Austin Federa. I am one of the three co-founders of D0ero. And D0ero is a bit of a different type of crypto infrastructure project.

So, what we're doing at at our core is we're trying to build a new internet. And we're trying to build a new internet from first principles, which is that the existing one is great for a lot of things and it's not great at a lot of other things. So if what you are trying to do is secure and synchronize Bitcoin, if you're looking at a roughly 21 transaction per second, 12 second block time L1 like Ethereum, it's great for all those use cases. But if we actually want to compete with traditional finance and traditional web 2 platforms, none of that actually runs on the public internet. You might think that Facebook and Google and Meta and all these companies run on the public internet.

They don't. They run in private data centers and they have some of the largest private fiber lines of anyone in the world. And so we started from first principles and we said if we're going to try and rebuild the internet suitable for fast execution environments uh whether that's L2s or Salana or other alt L1s um we really need to look at performance first and we have to build something that has decentralization and censorship resistance built in. but not over optimized for that the same way that L2s do not overoptimize for that. Uh and so this is where we came up with with double zero.

So what is this network at the end of the day? It's a deeply physical infrastructure network. And so uh this is a pretty famous gif of a shark biting a subc fiber cable. This is what the zero network runs on. It is a physical infrastructure protocol OSI layer 1 2 and three in the stack.

This is far below anything that you currently engage in in in blockchain with the exception of like ASIC miners on Bitcoin. Uh and so we're talking about these are physical zero devices. This is a high performance sort of off-the-shelf switch um made by Arista. And what we are building here is a physical fiber infrastructure network that scales across the globe, reaches um you know almost every market that we see significant amounts of blockchain infrastructure run in today and can do this much faster than the public internet. So what do we mean when we say faster than the public internet?

So the internet was not built for high performance systems. It's built for lowcost reach and it does a beautiful job at that. For 40 bucks a month in New York, you can get a gigabit of internet that can talk to any other server anywhere in the world. And that is a monumental achievement of capitalism, but it's come at the expense of performance. So what we have here is we have real data on the internet versus0ero.

This is from LA to Singapore, 24 hours of tracking. So what you see here, all of this variability in ping time, this is jitter. And jitter is a measurement of how unstable a connection is. So, we're all familiar with ping time. Ping time changes and drifts moment by moment and hour by hour.

And so, that's what's referred to as a systems jitter. And you kind of have this like large error band. And you can actually see the error uh the average latency jumps up significantly at this point in time. This is when the ISP just decided to route our connection over a different pipe, like a literally a different physical cable in the ocean that had worse performance than the thing we were on before. And that continues.

And you see these random spikes all the way up here and then down here. And all of this is a killer for high performance systems, right? If you build a high performance system, what you want is serialized data streamed into it consistently in the order that it needs to be processed. And everything about the public internet makes that extremely difficult. This blue line along the bottom is the zero network.

So not only is it lower latency, but it's got functionally zero jitter on it. And this is used by all of the world's tech companies, all the world's trading firms. Basically, anyone who's building anything serious in the traditional space uses private dedicated infrastructure because it's much more dependable and much more high performance than the public internet. Today in blockchain, nothing runs on it. So, we set out to build a network suitable to run all that.

So, this is the zero network today. This is fiber contributed by 15 independent contributors. So we build a software protocol, a blockchain protocol that allows independent contributors to mesh together fiber connectivity. So this represents about 8.8 terabs per second in total capacity, which is enough to run every blockchain in existence 10fold on the network today.

And the whole point of this is to allow folks to build systems that go faster and are more decentralized than the current high performance architectures of today. So if we go to the next slide here, this actually shows you what entity is contributing the fiber to the network. And you can see there's this quite dense network of multiple independent contributors. And so in many of the the top locations, you have a form of censorship resistance and decentralization. we have uh three different providers bringing multiple different pairs of fiber into this city.

And so even if one piece of fiber goes down, if one provider decides they're going to censor you for some reason, there's lots of other alternatives to deliver data. And this is really like the scale of infrastructure we need to be thinking on if we actually want to compete with traditional finance. Like today, we talk a lot in crypto about competing with traditional finance. We do a very poor job of actually competing with traditional finance. Price discovery happens almost exclusively on Binance for almost every asset in crypto.

Uh the most successful crypto products in Tradfi are stable coins, which are great. I love stable coins, but they're a completely non-competitive use case. Payments happen and payments have to happen, but there's no competition in payments. No one is trying to snipe you or take your liquidity or win a trade out over you. And anywhere where there's actually serious competition, traditional finance still runs circles around every execution layer in blockchain today.

We're we're excited about 10 millisecond block times or 10 millisecond pre-confirmations. And that's an eternity compared to the nancond scale traditional finance is operating on. And so if we want to actually start competing with these folks, we have to go back to the base layer and say what infrastructure is this running on? Most L2s today deploy their infrastructure in cloud providers. Almost every single one of them is built in a cloud.

That's also true of most crypto exchanges. Pretty much no traditional trady exchanges are run in the cloud because the cloud has huge performance hits and latency involved even within these systems. This is why you'll have 10 or 15 milliseconds of jitter just in an AWS data center. And so even when we sort of in crypto say, "Well, we're going to compromise on decentralization for other characteristics, we're doing it in a way that doesn't actually get you all the performance benefits of that." And so the zero network, it's it's live today.

We launched it in October. Um, and we have a bunch of real world data on how much faster it's performing. Uh, and we started off supporting just one network, but now we're at a place where we're able to support all types of blockchain systems as well as nonblockchain systems that are interested in running on the network. And I'll go into a few of the the other characteristics of it, but just to kind of go back like everything in blockchain basically happens at like layer 5, 6, and 7. This is way far further down in the stack.

And so when we're doing these optimizations, it comes from a very different place. So again, I said this is a blockchain protocol at the end of the day. There's core contributors to it, which are the people building the code. We have network contributors. This says 11 here.

We're actually up to 15 at this point. These are the entities that are actually contributing fiber to the network. So this is a sort of type of DPIN project. And then we have users which are validators, RPCs, traders, anyone who needs to communicate data faster than the traditional internet allows. But we can do better than that.

So we can use multiccast on networks like 0ero. Multiccast is a technology that's been standard in most traditional finance for 10 to 15 years at this point. You have run into it in your own lives. It is how Sonos gets one copy of the music from your phone to 10 speakers in your house. Chcast, AirPlay, they all work the same way.

This is hardware acceleration of packet replication. So in a uniccast world, every single destination a packet wants to get to is another data stream. This is why there's tax. Every additional node in a network requires more bandwidth for the person sending it. And so either you need a huge amount of bandwidth on the sender node or you have very complicated, very slow P2P data stream trees that propagate data all throughout a network.

And this is why if you're running flashbots or something like that, you are located next to the majority of the block builders so you have that data as quickly as possible. The beauty of multiccast is the network switches replicate the data on behalf of the users and so you can have multiple subscribers. It's very similar to like FM radio or something like that where the number of listeners does not increase the amount of bandwidth used by the system. Multiccast has been this is the way that like traditional exchanges distribute price feeds. This is the way that YouTube and Amazon and you know all these companies move data around their private networks and they have to get it to multiple locations.

This is so every single high frequency trading firm in the world manages their data movement. It's not possible on the public internet. So you need all the fallbacks. You need the ability to run through lib P2P or Turbine or you know Raptorcast or any type of state distribution system. But we can go faster than this.

And it's time we actually start taking this stuff seriously because if we want to compete with traditional finance, we can't be limited by running on technology that was fundamentally designed in the '9s for a very different internet structure than it is today. It's an achievement of blockchain that we're ready to do the thing that every major software company has done, which is grow up on the infrastructure layer. Facebook started off just building a software platform. Then they built their own servers. Then they built their own server racks.

Then they built their own data centers. And now they're building their own subc cables. And this is the journey of pretty much every successful technology company. Crypto is successful enough that it needs to start doing the same. We're already starting to build custom chips, right?

There's from AS6 was the first implementation of this, but we have zero knowledge prover chips. Now we're getting into hardware as an industry and the connectivity layer is going to be one of the massive unlocks that actually allows these blockchains not just to support 100,000 a million transactions per second, but to do it at speeds that are competitive with traditional finance, move price discovery off of Binance and back onto blockchain where it belongs and actually start competing. When we talked about tokenized equities, one of the questions I love is like why is there functionally no foreign currency trading on blockchain? We have stable coins in all these international currencies. Functionally 0% of the FX market trades on chain.

It's too slow, right? These these are these are small basis point trades where a little bit of price movement really matters. And today there's no blockchain system fast enough that any serious volume of FX trading takes place. That's a failure. That's like we've done all this incredibly great work in blockchain, but we're still not fast enough to really compete with traditional finance on trading.

So, where are we today? Uh, we launched our mainet beta, uh, which means there's still a few components that are more centralized than we'd like them to be for the long term. Um, we've rolled out a bunch of new features on the network and we're looking forward to a mainet coming up. But even with mainet beta which is a permissionless network that anyone can join. We have a bunch of examples on Salana of how we can speed things up.

So this is state propagation on Salana. This is a very Salana specific graphic but effectively on Salana blocks are 400 milliseconds and any alpha you can get on what's changed in the network is very important for trading. And so what we see is we have across the board hundreds of milliseconds of improvement even within data transmission within Europe. Let alone when we're talking about synchronizing data around the world where this increases even more by using multiccast jetto which is sort of a mev system on Salana they were able to reduce their bandwidth usage by 99.96% using multiccast and so if we look at scaling up these systems this type of bandwidth savings is really important this is the approximate difference between this pool of water and this swimming pool um so this is where we're at this is where we're starting to build things for high performance systems of the future And so, you know, at this point, we're talking with a number of L2s as well, a bunch of Oracle systems about adopting the zero network as an infrastructure layer.

If you're building stuff, if you'd like to talk, I'll be around afterwards. And uh you can learn more at.xyz. Thank you.