CoinGecko Podcast - Bitcoin & Cryptocurrency Insights

Building The Fastest, Lowest Latency Blockchain with Anatoly Yakovenko, Founder and CEO at Solana – Ep. 20

September 21, 2020 Bobby Ong Season 1 Episode 20
CoinGecko Podcast - Bitcoin & Cryptocurrency Insights
Building The Fastest, Lowest Latency Blockchain with Anatoly Yakovenko, Founder and CEO at Solana – Ep. 20
Show Notes Transcript

In this episode, Bobby Ong, co-founder of CoinGecko is joined by Anatoly Yakovenko, founder and CEO at Solana. Bobby interviewed Anatoly on Solana, its scalability, proof-of-history as well as Solana’s plans for Q4 of 2020 and beyond.

[00:00:02] Intro
Background on Solana
Scalability Trilemma problem
Solana blockchain size?
Other applications on Solana blockchain
Is Solana an Ethereum competitor?
Solana’s proof-of-stake with delegation validation model
Decentralized clock feature, called proof-of history
Solana’s plan for Q4 2020 and beyond
Where to follow Solana?

Quotes from the episode:

“If you go to our explorer, you can see that the network has done about 32 million blocks at this point and over 3 billion transactions. And the reason why we have 3 billion transactions already is because our smart contracts is so cheap and fast that we use it as the message bus for consensus.” [00:01:59]

“We don't have a limit on how big a block can be. There's a block every 400 milliseconds, and it isn't so much that there's a separate block. It's almost like a continuous chain.” [00:06:48]

“The hard part is obviously that [because] it is blockchain, everything has to be harder, right?” [00:25:52]


Solana -
CoinGecko -
Solana (SOL) on CoinGecko -

Social Media



Bobby Ong  (00:02):
Welcome to the CoinGecko podcast. I'm your host Bobby Ong. Each week we will be interviewing someone from the blockchain industry to learn more about this fast moving crypto currency economy. If this is your first time listening, then thanks for coming. The CoinGecko podcast is produced each week to help you stay ahead of the curve. Show notes can be found at I highly encourage you to join our newsletter where we send out top news in the crypto industry every Monday to Friday. Come back often and feel free to add the podcast to your favorite RSS feed or iTunes. You can also follow us on Twitter and Telegram at CoinGecko.

Bobby Ong  (00:39):
Welcome to the CoinGecko Podcast. For today’s episode, we would love to welcome Anatoly Yakovenko, founder and CEO at Solana. Solana is a high-speed single-layer blockchain using a network timestamp system called Proof-of-History. Before Solana, Anatoly led development of operating systems at Qualcomm, distributed systems at Mesosphere, and compression at Dropbox. He holds 2 patents for high performance Operating Systems protocols, was core kernel developer for BREW which powered every CDMA flip phone, and led development of tech that made VR/AR possible on Qualcomm phones. So welcome to the show, Anatoly.

Anatoly Yakovenko (01:11):
Great to connect!

Bobby Ong  (01:17):
Yeah, very happy that you can spend some time and join us and educate us all on Solana today.

Anatoly Yakovenko (01:23):
Yeah, likewise.

Bobby Ong  (01:23):
So I guess for the first question, right, maybe let's break it down in simple terms, like explain to our audience here at CoinGecko, what is Solana and how is it different from other Layer 1 blockchain?

Anatoly Yakovenko (01:35):
So Solana is a high performance Layer 1 and what we're good at is low latency and kind of high TPS so we can make transactions, get confirmed incredibly quickly. So under one and a half seconds. And the capacity is really high. We tested the system up to 50,000 TPS. We can do all of this without sharding and these are smart contract transactions. If you go to our explorer, you can see that network has done about 32 million blocks at this point and over 3 billion transactions. And the reason why we have 3 billion transactions already is because our smart contracts is so cheap and fast that we use it as the message bus for consensus. But all the nodes that operate in the network, right, when they vote, those votes are actually smart contract transactions. They pay for fee and those are valid transactions. So you see that the network is, you know, we kind of eat our own dog food. It's as cheap as you can possibly make it.

Bobby Ong  (02:38):
And how do you guys solve the Scalability Trilemma problem? I mean, Vitalik Buterin once said that blockchains can't have them all; decentralization, security, scalability and so have to choose at least one for trade-off. So between these three things, which one did Solana choose to have less off?

Anatoly Yakovenko (02:57):
So the trilemma only applies when the performance of the network needs to exceed the bandwidth that's available to the network. So what Vitalik didn't say is that there's ridiculous amount of bandwidth currently available. Some machines and that bandwidth doubles every two years. So if you look at like modern day hardware, like even your modern day 5G phones, the standard connectivity there is 1 gigabit. 1gigabit is a lot of data. You know, to give you like a measure of how many transactions per second can fit into 1 gigabit, it's somewhere between 500,000 to 700,000. So clearly, right, we don't have blockchains that have that kind of demand but we have hardware and phones and consumer hardware and consumer phones that have capacity for that kind of demand. So we don't really need to break the trilemma to solve blockchain scalability.

Bobby Ong  (03:56):
What do you mean with this example, this mobile phone 1 gigabit bandwidth example? Do you mean like the upload download time with the internet or like the amount of data that you can store?

Anatoly Yakovenko (04:07):
Yeah. So this is the upload and download time. So the 5G spec that's being deployed really quickly through Asia and somewhat in the US, literally the specification for 5G is 1 gigabit kind of activity while you're driving, you know, in the US, you want a 1 gigabit connection in China right, you see people driving at a hundred miles an hour or whatever that can maintain that connection while mobile. That means that standard connectivity in like, you know, your home or something like that, it's going to be between 1 gigabit to 10 gigabit fiber. Data centers are going to be 10 times that. So you can kind of like extrapolate from there. You already see like, you know, like a data plan in India, the costs for sending 1 GB of data is 9 cents. Okay. Well, what is the cost of sending 1 GB worth of blocks in Ethereum, right? It's like hundreds of millions of dollars.

Bobby Ong  (05:06):
So essentially what you're saying is that with this 1 gigabit 5G connectivity that we'll have in a few years time all across the world, we can upload and download one gigabyte of data every second assuming that 5G rolls out successfully and therefore it doesn't really matter to have like large blockchains and all.

Anatoly Yakovenko (05:22):
But this is us talking about technology that's in the future. What we have today, right, it's a hundred megabit and that's 50,000 TPS. So do you have, do you have demand for 50,000 TPS right now? Like in all of blockchain combined, right? We don't yet but data is already there. To give you like, kind of maybe a different perspective is that, like the only thing that a blockchain needs to do is to kind of take this transaction and replicate it across the world to as many validators or nodes or miners or full nodes that it needs to go to. So that's like 120 byte message. You can replicate it to 10,000 machines, you know, and like a hundred milliseconds. The cost of that is going to be 100th of a cent. And internet in Asia, in the US and China, globally, you can find anybody can do this, any consumer can do this, right? So that should be the bare costs for using blockchain. But because the systems that we have today were not designed to utilize hardware, right, to utilize the resources available to them, the cost is much, much higher.

Bobby Ong  (06:30):
And I'm just curious, right, so how large is the Solana blockchain these days? Like how many gigabytes is the blockchain and roughly on average, how large is the size of each block that is transmitted on the Solana blockchain?

Anatoly Yakovenko (06:48):
So the, we don't have a limit on the bytes that a block can be. There's a block every 400 milliseconds, and it isn't so much that there's a separate block. It's almost like a continuous chain. And the block producers rotated roughly every four slots. Each slot is 400 millisecond. So about 1.6 seconds. So the amount of data that goes in there is, doesn't have a cap. Right now, I think the network from consensus is processing for both consensus at Serum is at around 450 TPS. And that we see that just the raw ledger size is about 200 gigabytes. And I think with all the indexing and kinda like to do the rich query operations, it's about maybe one terabyte.

Bobby Ong  (07:32):
And recently you guys had a lot of headlines, made a lot of headlines mainly due to the FTX and Alameda guys. Alameda team, choosing to launch their new decentralized exchange, Serum on Solana. What was the impact that you guys see of Serum launching on Solana?

Anatoly Yakovenko (07:49):
The vision for Solana was really a vision of building the fastest possible, lowest latency, smart contracts extension. And fundamentally, if you remove the store of value use case, like if you don't have proof-of-work, before it's fundamentally price discovery because a smart contract platform is a financial system for programmable open. And because it's open and programmable by anyone in the world, everybody in the world is trying to extract free value out of it, right. They're trying to do an arbitrage on the information stored in the system that by definition is a price discovery engine. So our vision has always been to build the fastest, lowest latency one that we can physically possible. That's about kind of all the design decisions we've been making has been towards this goal. And I connected with Sam at this point about eight weeks ago. He always had a vision, they've been trying to build a decentralized exchange on basically for about a year.

Anatoly Yakovenko (08:50):
So they tried a bunch of different blockchains and always run into technical challenges and roadblocks. And I showed him our demo, Break. If you go to, you can go play with, play with the network. You literally smashed the keys to try to spam the network with transactions and they get all get cleared almost instantly. So they were actually impressed that this network was live and working and everything was, wasn't the white paper anymore. That it was actually functioning and get deployed. And like they tried a couple of things and like, basically things work in about four, we went from them knowing nothing about Solana to a demo in four weeks and then live in six. So, yeah.

Bobby Ong  (09:36):
Wow, they move really fast! In six weeks time, they got the whole DEX up and running.

Anatoly Yakovenko (09:40):
Yeah, it's already doing millions of dollars worth of volume per day. So that was really awesome because Sam had the same vision that finance is going to be decentralized, you know, sooner or later. And kind of like the first project that can make it decentralized, I think we'll probably capture a lot of value and fundamentally, like automatic market makers and those kinds of devices work great for low liquidity assets. But when you're dealing with high liquidity assets like Bitcoin and you're dealing with very large balance sheets, you need to be able to run an order book and sophisticated matching engines and take, and like builds kind of more sophisticated algorithms that just simply cannot be encoded in an automatic market maker. So this is been kind of a "our dream come true" to find this kind of lighthouse use case. And I always thought that this is going to happen in like, you know, three to four years from now because the space is still really nascent. Right. We don't see like the stuff that other folks have been building for decentralized exchanges isn't really like open and programmable. It's hard for me to take build up Binance on BNB Chain that competes with Binance. But everything about project Serum is open source. All the code is open source. Anyone can go and build and run their own market and attach on market makers to these things and actually have all the matching and execution happen in a decentralized censorship resistant way on chain.

Bobby Ong  (11:16):
Yeah, I completely agree. It's interesting how Sam has sort of demonstrated this potential that blockchain technology can offer on high scalability, blockchain. So besides the FTX guys, besides Sam and FTX guys building on Solana. Are there any other notable teams that are building on Solana blockchain or any other interesting dApps or applications in the chain?

Anatoly Yakovenko (11:36):
So before FTX and Serum, we already working with Ken and Terra. And these are micropayments use cases. And Terra is like, I think kind of a really interesting one because they are an established micropayments network or payments network, right? They have an application and customers and they're using Cosmos and Cosmos is not Ethereum. It's already much, much faster but it's still not fast enough because the traditional implementations of PBFT take about five to 10 seconds to come to a consensus. We're able to do this, like basically give them an optimistic conformation in 1.2 seconds, which is good enough for a visa, like point-of-sale transfer. So if you, the standard way of the pay, like anywhere where you have a customer experience is that the whole transaction has to finish in two and a half seconds from the point of view of the customer, right.

Anatoly Yakovenko (12:36):
And the blockchain part, there doesn't even have two and a half seconds to do. It has to be much, much faster than that. So I think we're really the only viable choice. And you see this, like Tether announced that they're supporting Solana, they're going to issue Tether on Solana and we're working with other stable coin providers as well. I think, you know, the boring stuff in crypto right now, like the payments, just business payments, point-of-sale payments. It's really, really boring use case. So how do I get users to pay each other? I think that's the exciting one because it has this really potential for really wide disruption across a lot of industries.

Bobby Ong  (13:15):
And do you view yourself, do you see Solana as mainly an Ethereum competitor? Or do you view Solana as a slightly different landscape?

Anatoly Yakovenko (13:23):
I personally like competition. I think it like breeds excellence, right? Like, so like, I don't know, I don't mind competition. I think it's fine to be in an Ethereum competitor. I think the, like I hate the term "Eth-killer" because that doesn't make any sense. Right? Like that's kind of a dumb way to look at it. Like a competition, I think is healthy especially because when you look at this technology, it is so complicated and there's so many different options and different ways you can build it that it's almost impossible for any team to get it right. We may get it right for specific use case. And I think this is where you start seeing differentiation. Maybe Solana is kind of the perfect chain for DEXes for financial applications, for payments. And maybe those are the ones that are the most valuable. Right. But there's definitely different approaches to design and build this stuff. And there's a lot to learn from each other.

Bobby Ong  (14:20):
Yeah. I definitely agree. The term Eth killer just assumes that there will be one blockchain that dominates it all. It may not necessarily be the case in the future. Maybe Ethereum will handle some sort of transactions and some of the blockchains handle some other niches and we see something else grow. So yeah, definitely competition is probably the right word. But I guess, do you, in terms of view of the other blockchain, there are many other Layer 1 chains that claims to have high throughput, like Tron or EOS, for example, like, do you view them as serious competitor or how are you guys better than these other high throughput Layer 1 blockchains?

Anatoly Yakovenko (14:57):
I guess, well, they are like version two, right? Like DPOS, Tron and EOS. And I don't know if you want to call Algorand or like those committee-based chains as well as kind of in a similar boat, the differentiation between, I mean, Sam and like a bunch of other guys that have tried to build exchanges on these chains, they couldn't get it to work. So there's some use cases that they simply don't are not scalable enough, right? This is where like outside of consensus, we do kind of a lot of operating system strikes and hardware tricks to actually accelerate the execution of the smart contracts. We have a very, like, you know, our smart contract language is Rust. It's harder to use than Solidity but it's fast, right? Like we made choices that are not developer friendly. We made choices for speed. And this is where they're paying off because things like at DEX really, really needs that extra few milliseconds.

Anatoly Yakovenko (15:54):
I don't know if that's like the difference there because we were so fundamentally focused on latency and performance that that actually opens up new use cases that weren't before. And that there isn't a clear differentiator between now Tron, Neo, Algorand and Cosmos. Although, you know, like there's ecosystem springing up around things like Cosmos that are not performance specific, right. This is kind of all about interoperability and similar to Polkadot, right. There's its own ecosystem, which is not as nobody there cares to go from five seconds to one. Although I'm sure the users will be happy but this isn't their main focus.

Bobby Ong  (16:37):
And do you think like the trade-off of using Rust and not having it very developer friendly will be an impediment to your growth because I think one of the reasons Ethereum grew to where it is today, it's got a very vibrant developer community. Most of the developers who are interested in blockchain technology inevitably just ends up learning Solidity and Ethereum. And I think one of the key reasons or key factors in growing a blockchain is to have a very vibrant developer community. So just want to hear your thoughts about that.

Anatoly Yakovenko (17:05):
So the interesting thing is that Solidity is actually not easy.

Bobby Ong  (17:08):
Of course!

Anatoly Yakovenko (17:08):
And this is where like, people, I think get it wrong is that they think that if you make it easy, you're actually solving a problem. Engineers are like trained to solve hard problems, right? So making it easier for engineers isn't actually what you want. You actually want to build something that an engineer can use to solve a problem that they have. Not like, they don't really care so much about easier tools because they've all coded in Solidity, right, two years ago when it was barely usable by anyone, right. Not even by the people that wrote it, right, because there was, it was impossible to build, you know, these applications anywhere else. So like the key for us really isn't so much like, you know, we're lucky we're building on Rust. Rust is a modern language, Facebook, Google, Apple, Amazon, they all publish open source software written Rust.

Anatoly Yakovenko (18:07):
The Rust developer community is probably 10 times bigger than Ethereum, just because it's like a widely used language, right. But it's not JavaScript. And the problem we're solving, isn't the developer experience of using Rust versus JavaScript. That's the low latency thing. Can you like, it doesn't matter what tools you have, right, if you can't get the problem that you want solved, makes no difference. But if you see that, like there's a business opportunity or like your, the idea, right, the innovator has, and only be done on this network, then this is what they're going to do. This is what they're going to use. So like we're making kind of a bigger bet but it's like a saying, like in web development that like, you know, "Engineers will like eat glass for latency". There's a certain number of, certain kinds of applications that just like need the fastest possible response times. I personally believe that those applications are going to be the ones that capture the most value because those are the human interactive applications. As long as, you know, if you're building for people, right, for humans and stuff just gets response, faster responses, they will not want to switch to anything slower.

Bobby Ong  (19:22):
It's interesting point of view. And based on my understanding, Solana uses a delegated proof-of-stake validation model.

Anatoly Yakovenko (19:28):
It's proof of stake with delegation, not DPOS on the other blockchains. Yeah. It's just proof-of-stake, but the contracts are programmable. So you can kind of like assign and change delegations. We actually have this, if you go to like our Testnet called TDS, kind of like the Kusama with Polkadot, we're running these like staking agents that are forcing the stakeholders to maximize censorship resistance. And if you look at that network, it's probably, I would bet that this is the most censorship resistant network on the planet right now. Because it takes like over a hundred validators to get to 33% of the stake. So if you look at every other proof-of-stake network out there takes like five or 10, right? Like somewhere between five or 10.

Bobby Ong  (20:17):
So it's proof-of-stake but they are validators, but it's not exactly delegated proof-of-stake.

Anatoly Yakovenko (20:26):
So like DPOS, like Tron or EOS usually implies that there's some group of elected elevators who control consensus. We don't have any limits on the stake size or the number of validators that can participate in consensus. So there's no limits there. If you have the smallest amount of stake, right, like let's say you have, you know, one 10 thousands of the stake, weight, this, you will get one block out of every 10,000 to produce. So there isn't like any upper and lower bound.

Bobby Ong  (20:57):
And anybody can be a validator, right? I believe.

Anatoly Yakovenko (21:00):

Bobby Ong  (21:01):
Why do you choose proof-of-stake with delegation instead of proof-of-work or just plain vanilla proof-of-stake?

Anatoly Yakovenko (21:07):
I mean, we are basically choosing vanilla proof-of-stake but if you look at Ethereum, they have limits. You have to have at least 32 Eth to stay for Ethereum. So I didn't want any limits because I think they're kind of cheesy. It means that you're like, you're avoiding a hard problem. Right? Like we really wanted to make sure that the problem we're solving is the right problem. Right. So this hard, right. It's hard to make sure that you can actually scale the network to the largest number of participants in consensus. So fundamentally, like if we're talking about consensus and decentralization, the key part to all of these systems that is really, really hard to fake is the minimum number of machines that's necessary for an attacker to like take, get physical access to help the network, like the live with this threshold. Right. So in EOS it's eight. You look at Cosmos or Tezos, it's actually lower than that. Six or seven.

Bobby Ong  (22:06):
Oh, wow. Okay. I thought you were gonna say 20, but like six or seven. Okay.

Anatoly Yakovenko (22:11):
Right? Because this is the number of validator, the smallest number of validators that it takes to get to 33%, right. As soon as you control the minimum, kind of the super minority or the liveliness threshold of the network, you can halt it, right. So you can sensor the network. So this is like the hardest number to kind of fake. You can get, you know, you can build proof-of-stake networks or delegated proof-of-stake networks with like a thousand full nodes, but those full nodes with no stake do nothing, they're useless. Right. They don't really add any value to the actual censorship resistant part. So, and the reason for that is because it's really, really hard to make those systems performance when that set gets really, really large, because now what you're talking about is synchronizing all of those machines on every block, right?

Anatoly Yakovenko (23:01):
Not just a small subset of those machines and I'm propagating it kind of like later to the tail end, but actually having the network come to full consensus, right? Like, and that's a much harder problem. We've actually solved this. If you look at like our TDS for we're experimenting and like testing these staking agents, like it runs as fast as main node with like, you know, effectively over a hundred nodes to get to the liveness threshold. So there's a huge difference between China get physical access to eight machines, right? Or less than that, then a hundred independently owned validators, like globally.

Bobby Ong  (23:37):
And you guys introduced this decentralized clock feature in Solana, called proof-of-history. Maybe tell us a little bit more about this and why is it a problem in the first place?

Anatoly Yakovenko (23:49):
Like the whole proof-of-history was the reason why Solana started. To give you guys an idea of why time is so important, like imagine, you know, first radio towers were built. You have two radio towers transmit at the same time, over the same frequency, those electromagnetic waves that propagate through the air because they're at the same frequency, they interfere with each other and you get noise. Information can't pass, right? So the first protocol anyone's ever built is it gave each tower, a clock, a synchronized clock and they alternated by time. So one tower transmit, it's like a second one, then tower two transmit on even, right? And the more synchronized your clock is, right, but more accurate that synchronization, the faster you can rotate them. So modern day cell phones, like I think it's somewhere between 20 milliseconds or five milliseconds for CDMA.

Anatoly Yakovenko (24:41):
I don't even remember at this point. But those clocks are super, super synchronized now. And that means that you can have many number of transmitters per second, right? Because you can not quickly alternate. So this protocol is called time division, multiple access. Silly name, right? But it is actually the exact same problem that you have in proof-of-work networks with block producers. So if you look at like Bitcoin, right, the reason why you have 10 minute blocks is because Bitcoin is obsessively trying to minimize the likelihood of two block producers transmitting a block at the same time. So this difficulty adjustment is there to slow down the network, such that it's extremely unlikely that this one megabyte block gets produced at the same time by two different transmitters. And then, then the network is confused and doesn't know which, which is actually the real chain, right, which works. So that confusion exactly the same as the noise and radio. So if you have a source of time before consensus, like a global point of reference that everybody can agree on, you can start coordinating with that point of reference. The hard part is obviously is that blockchain, everything has to be harder, right?

Bobby Ong  (25:56):
Otherwise where's the fun right?

Anatoly Yakovenko (25:56):
Yeah, exactly. Right? Like otherwise what's the point, right? So it needs to be trustless and therefore not relying on like a global clock like Ethereum and CDMA or on atomic clock, like Google uses for Bigtable or SPAN or whatever they call it these days. So we got lucky, but you know, computer science problems, you don't solve them. You just get lucky. We got totally lucky, right. That they were able to figure this out. And I like in late 2017, I was thinking about proof-of-work.

Anatoly Yakovenko (26:33):
And I had this kind of Eureka moment that we can use the same SHA 256 algorithm that you use in proof-of-work but you said recursively and you can generate a piece of data that represents time passing because it is represent a single core, a single CPU core somewhere in the world, spinning for certain amount of like, cycles. And because TSCM and Intel and every kind of global foundries are basically at the same process layer. And the differences between them are about like at best 10 to 20%. You can get a really good estimate of how much time, how much real time has been spent. And it's really, really hard for an attacker to make that 10 times faster. It'll cost them billions of dollars basically because they have to build a better app, right. And fabrication process is really, really hard to optimize. So that's kind of like where we got lucky is that we were able to solve this computer science problem. The Rust is just more standard database, you know, operating system kind of optimizations that, you know, myself and most of the ex-Qualcomm folks that joined, you know, we've worked on those problems like, you know, for a decade.

Bobby Ong  (27:49):
So basically what you're saying is that every machine or every node runs like SHA-256 trying to solve a puzzle. And the time it takes to solve that puzzle is sort of how you back-calculate the clock and set a clock.

Anatoly Yakovenko (28:03):
Yeah. So they're all basically all run a single core in a loop and that loop generates some data. And we're able to use that data and kind of glued together to build almost like a water clock. So you can think of it as like water dripping. It's like this data keeps in a similar way that a blockchain everybody [inaudible]. They kind of do a similar thing. And that data structure together creates some global point of reference in time.

Bobby Ong  (28:29):
But like there could be a time in the future, say like five or 10 years time, where fabrication gets super efficient and then it becomes like, I don't know, five times faster to solve the puzzle, for example.

Anatoly Yakovenko (28:39):
Sure. But like the way that the application process gets better is like TSMC goes from like, you know, five nanometers to three nanometers. Then that speed up is about 25%. It's never like at a 10X. I mean, who knows, right?

Bobby Ong  (28:58):
I don't know, I'm not a chip designer so it's not my area. It's good insights, nevertheless.

Anatoly Yakovenko (29:01):
So even if there is like a 10x speed up, it's not like a way to double spend or any, or it's not an attack on safety. What that does is it allows the attacker to really slow the network down. So it kind of like all of the work of an engineer is to take the best case and make it the average case most of the time. Literally like, right. So like if an attacker has a much, much faster clock than the network really behaves like DBFT. So we get these like wonky five to 10 seconds confirmation times. Because now you have multiple block producers that are producing blocks at the same time. And everyone's confused which one to pick, which is fine.

Bobby Ong  (29:40):
So how do you guys, like, you know, you gave the Bitcoin example, like one block is produced every 10 minutes. So the chances of two blocks being produced at the same time is quite rare, but like yours, you have like blocks are produce every 400 milliseconds, I believe. And so there's always cases where people have competing blocks. So how do you solve this? I mean, in Bitcoin is the longest chain solution, like, I don't know if it's the same for Solana.

Anatoly Yakovenko (30:02):
Yeah. Yeah. So like Bitcoin has a heaviest chain. We use something more like Tendermint-like DBFT. Although optimized and kind of like, I don't know if you've heard of Nagle's algorithm.

Bobby Ong  (30:18):
No, please simplify it.

Anatoly Yakovenko (30:18):
That's also like one of the like kind of go to engineering tricks is when somebody gives you tasks, you don't complete one task at a time. You get like, give me as many tasks as you can and then I'll do all of them at once. And making that kind of number variable, it's called the sliding window. And that's basically kind of how one way that we optimize it. But to answer your first question is that even though we're producing blocks every 400 milliseconds, because we have this proof-of-history clock, we never seek collisions because the block producers have to generate this data structure and they can't fake it. So they actually are forced to delay for a milliseconds. So block producer one produces a block right now. And the second one has to spin for 400 milliseconds to get to their alloted time. And this is what allows the network to kind of seamlessly transmit a block effective continuously. You can almost think of it as just one giant block that have, you know, alternating producers.

Bobby Ong  (31:20):
I think I have a better idea now. So let me ask you some maybe simpler questions now after this. 2020 has been a crazy year. What's the plan for Solana as we gear up, I mean, you guys achieve a lot in 2020 with the FTX guys on board with Serum. What's the plan for Solana as we gear up towards Q4, maybe some idea of 2021, maybe?

Anatoly Yakovenko (31:42):
I think like in the kind of the medium term, we have a bunch of projects in the pipeline, some of which have been announced already, but like some of which we haven't. And they're focused on a broad range of use cases, financial ones, for sure. And more like even boring finance, like payments and those things but also like consumer facing applications where we see potential for like onboarding, you know, millions and tens of millions of crypto users. Like right now we see, like, if you go look at like, you know, the top dApps for Ethereum, you look at their total user bases, their sub are 100,000 right? Now the Uniswap airdrop went to 60,000 people, right? Like, which is amazing that they have something this sophisticated and complicated and like hard to understand for an average person actually have 60,000 users. That is, I don't want to like underestimate that accomplishment because that is awesome.

Anatoly Yakovenko (32:42):
Right. But crypto is still like really, really small. And I think there's a lot of potential to get, you know, millions of users doing something on crypto related on the web, especially around social networks and kind of like content and like more interesting kind of like consumer friendly applications. So that's kind of like, we've been working on now but also the fundamental part is that like, I think project Serum is a lighthouse application. Sam is awesome. The FTX team is like, I love working with them because they love turning coffee into code. They will just like churn stuff out until it's done. So I kind of love that energy. That's very kind of part of our culture to just like crank stuff out until it works. So with them, like, I think we've seen since kind of the start of Serum, like this was almost like a black hole of projects coming into Solana and building on top of Solana.

Anatoly Yakovenko (33:40):
And we are starting to see people just announced like wallets and applications and kind of like full demos and applications without even talking to us first. That to me has been like really, really cool. But the cool thing about Serum is that it's new, right? It's a totally different thing. It's not Uniswap. It's got a totally different style of way to do trade and financing, to do like decentralized finance and the kinds of applications you can build with it are going to be unique and different. And the same way that you can think of like Aave and Compound better part of the Uniswap ecosystem. Because a lot of folks, they lend and, you know, borrow from Aave, they use Uniswap right to their bar collateral. Right now you can build lending protocols that plug into a spot market that doesn't require an Oracle. That is local on that chain.

Anatoly Yakovenko (34:34):
But you get atomic guarantees about not just the price, but like what does the fill, like what, how what's the depth of the order book, right? Like these things become much more interesting because I think they will allow much larger volumes to be actually running in DeFi applications. Not like recursive TVL style, like valuations, but like, but like Serum announced that JumpTrader is going to be market-making on Serum on a decentralized thing. Right. They have an absurd balance sheet. To take like a billion dollars and put it at risk at an order book, right? Like you need like the stuff to work. Right? And this is, this is what we're opening up. And to me, this is like the most exciting thing it could be working on because it's really pushing the network and everything that we want to do on the engineering side, just kind of like napalm, right? Like we want to get the block times to, you know, theoretically 80 milliseconds, but let's get it to like 200 first. Like we're going to be the first chain with native execution. So like the byte code translate it to machine code. So like a Git or, you know, ahead of time execution, like you see on like Android or iOS, where you're actually like pretty close to being done with that. So like stuff like that is like, I think basically I think by the time Eth 2.0 Launches or any of our competitors actually get to anything interesting, we're going to be in V2 of like a hyper speed.

Bobby Ong  (36:14):
Interesting. Very interesting plans. And I definitely agree. I think one of the key factors for having DeFi running on more mainstream applications would be that if a centralized exchange market maker can move their market making to a decentralized exchange and suffer no performance loss, we know decentralized blockchain, decentralized applications have achieved its full potential. Because if, as long as there is a lag and a DEXes relative to a centralized exchange, then there's always the argument that central exchanges are more efficient. But once that is no longer the case, then there's really no reason why everything shouldn't move to a decentralized exchanges after that.

Anatoly Yakovenko (36:51):
So the reason like, even when folks talk about high frequency trading, the DEX right, this global agenda of price discovery, it only needs to be as fast as the news. So like you have like, imagine a network, 10,000 machines, all fully connected fiber, right? 80 milliseconds round-trip time from hear to the others type of world. Some, some event happens newsworthy, right? And this data goes into this state machine. It's propagated in 80 milliseconds across the world. That propagation is going to be as fast as that data moving through Bloomberg terminal. And the high frequency trading where, you know, you hear people talk about sub one millisecond, right? Like microsecond trading. That's actually not trading. That is trading against the news. It's trading and that's occurring in a local market of orders coming into the queue against like the real market, which is like the global price discovery. So like HFT system running on top of Solana, to me like this is the end game, right. We are like the price discovery engine for the world. All of like news and real assets occur here, right. And then you have local markets that are doing like their fancy HFT, you know, as fast as they can, like whatever they want to do there, sieve signals from noise, and then effectively those markets can pay traders to push their orders through there because they're optimizing on this like really, really small timing.

Bobby Ong  (38:18):
Yeah. Very, very cool. I mean, we have HFT traders coming on Solana. That will be like, "Whoa, you live in a totally different world by then". All right. I guess one last question, right. If someone's really interested to find out and learn more about Solana, where's the best place to jump in and have discussions and learn more?

Anatoly Yakovenko (38:34):
Yes. So go to There's a discord there, Go talk to us. There's docs. If you're an engineer, go, tons of documentation. If you want to just like see how fast the network is, go to By the way, all of these things are So like probably just type in Yeah. It should be fairly easy.

Bobby Ong  (39:02):
That sounds very interesting. I had a great time talking to you. I learned a lot about Solana, asked some hard questions on how you achieve scalability, decentralization and so on. So I'm very glad that you answered and educate us here at CoinGecko, on this podcast. Thank you very much.

Anatoly Yakovenko (39:17):
For sure. I love CoinGecko. It's awesome. You guys are super easy to like talk to and communicate in terms of just like interfacing with you. It's been a pleasure. So like you guys are good folks for the community.

Bobby Ong  (39:33):
Yeah. We do our best. We know that crypto is all about community. We want to help grow the community. So everything that we do is always community first. And if it means like working for hours and all, to work with the team closely to get our information up and correct, we have to do that because we know that a lot of millions of people rely on our information on a site and we need to get it accurate for the rest of the community, the crypto community, to know about happening in the industry. So, yeah. Yeah. Alright. Thank you very much for joining us on the podcast.

Anatoly Yakovenko (40:00):
Yeah. Awesome.

Bobby Ong  (40:02):
All right, that wraps up the show. Thank you for listening to the CoinGecko podcast with Bobby. If you like our show and want to know more, check out or please leave us a review on iTunes. If you have any feedback, do drop us an email at Join us for more next week. See ya.

Disclaimer (40:21):
This podcast is provided as part of the overall information on cryptocurrency contained on our website is where your general information only and does not howsoever constitute any endorsement, financial or investment advice nor any solicitation or offer of securities or other financial instruments, coin gecko, and the podcast presenter makes no warranties implied or express of any kind in relation to this podcast, including without limitation the accuracy and updatedness of its content. All opinions and recommendations there in the podcast are based on the personal opinion of the presenter. Please conduct your own research and procure professional advice. Should you at your own risk, decide to howsoever invest or trade in relation to the content contained in the podcast.