Storage Developer Conference - #121: Storage Applications in Blockchain

Episode Date: March 17, 2020

Hello, everybody. Mark Carlson here, SNEA Technical Council Co-Chair. Welcome to the SDC Podcast. Every week, the SDC Podcast presents important technical topics to the storage developer community. Each episode is hand-selected by the SNEA Technical Council from the presentations at our annual Storage Developer Conference. The link to the slides is available in the show notes at snea.org slash podcasts. You are listening to STC Podcast, episode 121. Today's presentation will be on storage applications and blockchain. Hello, can you guys hear me? storage devices, and blockchain application. On that note, I'll pass it to Daniel. Hello. Can you guys hear me?

All right. Who here is familiar with what a blockchain actually is? Okay, great. Who knows where the actual security in a blockchain comes from? Public validation. So we're going to talk about some of these things and hit some highlights about what an actual blockchain is, different consensus protocols, as Olga pointed out, as well as going over some of the demystification of the new buzzword

of modern day, I guess you could say. So we're going to go ahead and get kick-started. So a blockchain is nothing more than a distributed database. When you remove public validations, that is all it is. The validations are the cryptographic functions that verify transactions that are being stored in said database. That is where the security in a blockchain comes from. So the fact is, if you are familiar with what Bitcoin does, meaning that everyone is performing the math to prove that it happened,

every peer and every miner has to agree on this. Otherwise, there's a contestant and the transaction is usually dropped out of the memory pool or dropped out of the chain altogether before it's ever applied. Blockchain as a technology as a whole has been, we'll say, perverted a little bit and has been pandered towards telling people everything from it'll cure cancer to it'll cure poverty, which it does none of those things from a fundamental standpoint. And we're going to go to some in-depths about what that actually means for use when you're talking about storage and where there are consensus protocols such as proof of work, what the differences are and where it applies to storage,

and how the new consensus protocols are actually starting to utilize storage versus heavy draw power from CPU, GPU, and ASICs. Current state of the blockchain technology is we have a lot of people are starting to use DApps. You may have heard that word. It really just means an application that may utilize a blockchain. It doesn't necessarily mean that everything needs to be on the blockchain in regards to it. It literally just means that it makes a transaction to store some kind of state. So if you're already familiar with writing applications,

then this part should be pretty simple to understand. Everything that you would insert to your database essentially is a blockchain transaction. If you want to break it down to the simplest, lowest common denominator, that's the one way that they're both in parallel. Now, there are current reports of 2017 and 2018 that a lot of the ICOs that were DApps,

a lot of these other things have literally been nothing more than scams. Some of them have failed to raise funds. Some of them have also failed to deliver some of their promises, which, if you're familiar with Silicon Valley, that happens a lot with early-stage startups as well. The only difference is now there's a ledger of who raised what kind of money. Now, when we start talking about where the current environment is, there's a lot of talk about scalability when it comes to blockchain. Most of this is referenced to where Bitcoin has failed to scale.

And we'll go over that in a little bit when we start talking about the consensus and why it matters specifically related to storage all right so many larger corporations as you can see you're becoming more interested think IBM Hyperledger and stellar various projects, all of them stand to try to give this idea of using the blockchain as a service. Now, a blockchain inherently is nothing more than a distributed database. Now, once you start talking about the introduction of what is called cryptocurrency, or currency that's generated as a byproduct of the sustainment of the blockchain, by the way. So to give you some perspective there, everyone that's mining is mining in order to validate the transactions,

and those transactions as a reward are given as a block reward or a transaction fee to those that get the answer to the puzzle. This is the way of incentivization. But the problem is most of these companies are removing the larger scope of the public peer validation, which is inherently perverting the security of a blockchain and thus making it weaker. However, they're promising you higher throughput. So realistically, though, at a point in which you sacrifice speed and throughput for the security in which you were promised, at what point is the database actually more secure than a blockchain? And we'll go over that later.

A lot of companies and individuals, states, governments, pharmaceuticals, everything from supply chain to anything that you need public or more than yourself looking at the data that's being verified as an actual use case for where a blockchain stands to shine. However, you have to remember, just like with any information put on the internet, once it's there, it's there. Now, whether or not you choose to put that information in a way that you and your partners choose

to have some public-private key pair exchange in order to view it, that may change and allow you to actually operate publicly being more secure. One of the things that you're starting to see also is large-scale farming operations, both on the cannabis side and the vineyards are starting to roll out the technology to follow from the second your grapes or your cannabis is planted to where the weather reporting and everything is starting to be stored in these records so that the individual end user who may have purchased said supplies actually

knows where everything happened. However, this can also be done with a database. So let's not put too much on what it is. Now, as we can see here, UBS, Microsoft, PwC, Wells Fargo, plenty of people are starting to adopt blockchain. As we start to see the growth, huge amounts of growth started in the 2016-2017 timeframe. A lot of the stuff you see in 2018 and 2019 as far as the number of companies, and these are the estimated bank spending on the tech,

a lot of that is just now being the return on what those people were investing in in 2017 and earlier. Let me ask you, does anyone know what other companies were missed and who else is right now adopting blockchain technology?

I think Deloitte and Touch does some, right? Yeah, Deloitte does. Oracle. I mean, everybody's a dog. Yeah, Oracle has their own service. Amazon. Amazon? Yeah, Amazon offers blockchain as a service through AWS.

SAP. I don't. Yeah, SAP. So who else? Anyone? Come on back. I mean, a lot of companies are, yeah. Yeah.

All right. So one of the things that often people talk about is the limitations of the blockchain. And this is where we're going to talk about that, as I have already mentioned it. So one of the problems right now is that with Bitcoin, Visa, and other blockchains, is that there's this magic number everyone wants to get to, which is about 1,700 transactions per second. Now, when you start thinking about that on a global scale, many of us have applications that do way more TPS when we start talking about rights to a database than what Visa does.

When you start looking at breaking that down versus a blockchain versus a business, there seems to be a lot of misunderstanding. And so when you start talking about Bitcoin, Bitcoin isn't ran by a business. Neither is Burst or any other blockchain that we're going to be talking about specifically. And the fact that these are individuals who run software at home or on lease servers that they have in a data center or wherever it is they may choose to. And from there, what they're doing is they're processing the transactions. So when you start thinking about how do you keep that system intact without compromising the user,

it becomes very hard to think about scaling. So in 2009, when I first got involved with Bitcoin, you know, it was like one to two, three, I think it was 1.5 TPS as basically what we were lucky to get. But at the time, it made sense. There was nobody using it. As people began to use a blockchain and as people began to use the technology such as Bitcoin, we started seeing limitations where now you can see there's like 4.6 a second.

I think at its peak, it can hit 7. There have been times when, you know, the chain itself was performing better, but there was less transactions. So as we see it now, you know, roughly you have 7 transactions per second, but now that's not going to immediately go through from user A to B.

You have to wait on confirmations. The confirmations are the validations that are done by the peers. That takes time. So when we start looking at where these systems are at, they're very much in their infancy as far as where things are going to go, because the legacy finance system has already been established in a way of transferring data. At a point, all these transactions literally are, it's just data. So this is a very rough draft about how a transaction is sent through, and I'm going to walk you through it so it makes a little

more sense. So we just use the B logo for versus as representation, but basically a user will send a transaction or request, it's broadcasted to every peer on the network. So that means every machine that's running that software gets it in what's called a mempool. Every transaction is stored in a mempool until it's validated. Transactions are approved by doing the consensus. Once it's done, it's added and solidified in the blockchain and cannot be rolled back unless there's a 51% attack, but we'll go over that later.

But basically, this is a very simple outline of how a transaction goes from user to actually being added into the chain. And the approval part is the actual consensus, which we're about to go in detail about the different types of consensuses that there are. And one of the consensuses will make a lot of sense as to why we're talking about it here with the storage. So we have proof of work. We have proof of stake.

Delegated proof of stake. Proof of space and proof of capacity. Now those last two might have something very interesting to do with storage. And that's why we wanted to talk. So proof of work is the first trustless consensus. What that means, and this is the consensus that Bitcoin uses, what that means is that any participant or anyone that wants to participate doesn't need to ask permission to be able to participate to validate transactions.

Meaning I don't need to know you, you don't need to know me, we can do the math. Whoever gets the puzzle, we agree. Whoever won based on the signatures. And this allows for everyone to be able to participate without having to ask these permissions. And because of this, this allows it to be trustless. Meaning I don't need to know you. You don't need to know me. You don't need to know my background. You don't need to know what kind of computer I'm running or operating system.

But we can still participate in the same network. Now, because of this, there's a game theory that sets into place, which is one CPU, one vote, which was set forth by Satoshi Nakamoto in 2009 in the Bitcoin white paper. Actually, August 31st, 2008 in the Bitcoin white paper. But basically, machines or users who had more powerful machines

would be able to do more hashes, thus being able to have a higher chance of getting the reward for the actual block that's proposed. Now, the individuals who are participating by actually performing the SHA-256 are what's called miners. So if you hear me refer to a miner, I am talking about an individual who is running the consensus. Now, that may change from proof of work to proof of state to delegated proof of state or proof of capacity or the others. Now, literally, mining can get kind of muddy because of the way some of the other protocols work,

which we're going to talk about. And then we're going to talk about where proof of work and proof of capacity and proof of space are aligned with being a trustless consensus. Now, along with what goes on in proof of work, a lot of people are starting to hear more and more about ASICs. So there are a large number of ASICs that do proof of work. Bitmain, being a very large company that was producing tons of ASICs, Antminer, the rest. Litecoin and others have tried their best to prevent it from being ASIC-fied.

But Monero has been pretty good as they fork their code base every six months to prevent ASICs from going into circulation to be able to gamify the mining process. This allows for it to remain fair. So when we start talking about different types of consensus, there are a few that have trade-offs, and we're about to discuss one of the big trade-offs, proof of stake. So unlike proof of work, it is not necessarily trustless. Meaning that in some cases there's a theory, and you may see it out there, called masternodes. You may have seen that on some advertisement on Google or something. And basically all that means is that if you wanted to be a masternode, you're essentially validating the transactions on that network. But people are staking or promising their coins

for a percentage of your take of those validations. When you start talking about proof of stake in certain coins, such as Dash, that allows the user to stake their own coins to be able to validate. However, not all systems work this way. Now, proof of stake versus proof of work doesn't utilize as cryptographically sound block generation methods so there are some issues there where there are some people that are doing some dirty things and we have seen that on the public eye and has been on cnn

but the difference is that there's a nomenclature difference for referring to the miners as forgers versus like an actual miner. Now, the difference between proof of stake and proof of work is that in a proof of stake system, you may see 11,000 to 20,000 transactions per second versus proof of work, where you may only see, I think right now, some of the higher performing throughput coins are doing, I think, 100 TPS with proof of work. And at that time, though, you still are compromising the user. But in this case, what they're doing in proof of stake is basically instead of everybody participating, only the ones who have met the requirement to be a masternode or a validator is allowed to do so. Now, the currency can be inflated or deflated as they choose,

as they may print more tokens. So versus Bitcoin or other coins having a fixed supply, proof of stake is literally nothing more than an arbitrary number and a piece of code that they change. Now, the same can be said about Bitcoin, but the only difference there is that when you make that change in Bitcoin, you have to get consensus of everyone. And proof of stake models, Bitcoin, you have to get consensus of everyone in proof-of-stake models.

You don't have to. So lastly, we're going to talk one last thing on proof-of-stake and move to DPoS, and then we're moving to the storage stuff. So Ethereum is going to be moving to proof-of-stake. Stellar and EO are already using the consensus, and you may have heard of them as the DApp movement and things like that.

Plus, they have a large marketing team. All right, so delegated proof-of proof of stake is an interesting proofing model. So basically it's really parallel to the old banking system. It's you vote for a producer and they validate on your behalf. So you still stake coins to this group or an organization. And EOS being one of the largest DPoS coins, recently took a lot of flack for certain things in regards to corruption inside the voting system. And so basically, if you're a participant in that network,

you can choose who's going to validate your blocks. Now, because you can choose who's going to validate your blocks, that means you have some say so to some level but you're again still allowing them to validate that and not allowing you to now uh also included in the dpos system is a much faster throughput because typically the people who are involved with it are businesses. So they're very large businesses with very large bandwidth pipes and are literally essentially creating a distributed mining service for individuals for these types of consensus protocols.

Let's see. But yeah, blockchain is such as the US. Cardona, which is a bigger one that people know of, and Tron use this consensus. And I mean, quite literally, it's the same as what you would see in a typical blockchain layout, except instead of you being able to validate, someone else is doing it for you. All right, so proof of space. Now we're talking about storage and the beginning of using storage concepts.

Our storage devices is actually a consensus protocol or being part of a consensus protocol so right now proof of space some of the concepts that will be used will be chia and a coin named spaceman now chia has not actually produced anything yet spaceman is still a conceptual idea the poc or proof of capacity blockchains are often confused with proof of space, but there is a different timing mechanism that was incorporated in the papers. But they were later not implemented from certain coins, mainly because proof of space lacked in certain validations. Most of them related to VFD, which I think Chia is actively working on fixing. The verification was that, you know, you needed a way of verifying that a certain size was a certain size. I mean, this was a two-terabyte drive.

How do I prove, mathematically, this is a two-terabyte drive without compromising the space that's on the drive, right? So this is where proof of capacity and proof of space we're attempting to work out a solution to be viable as a validation process whereas proof of capacity versus proof of space proof of capacity is actually in use meaning that it's like this little device here and just like this hard drive right here i am currently mining and i am reading from my allocated plots, and I'll go over that process here in just a moment. Proof of capacity allows anyone,

just like proof of work, to mine. You just need to be able to have storage. Now, you validate that storage by generation of what's called a plot. Now, a plot is a cryptographic function output of a Shebol 256, which is a hashing algorithm. Now, what this does is it allows for me to fill up the entire size,

such as this, which is a, I filled up a terabyte on this, and I'll be showing, demonstrating the mining here in just a moment. I'll actually show you while we're here. Give me one second. So in the background here, we are actually mining off of this one terabyte as we speak. So I wanted to, I'll come back to that and answer some questions when we get a little further along.

So basically... And to simplify it, you can say it's really the... Yes, so to make it real simple, a SHA-256 is never actually written to anything when Bitcoin does its proof of work. The Shabal-256 requires an actual write to a disk. So because it requires the actual write, it's the I.O. throughput that prevents it

from being ASIC-fied right now. Current rate, even with NVMe, and we'll be showing charts of that, it still takes a significant amount of time to be able to actually plot the drives, rate even with NVMe, and we'll be showing charts of that. It still takes a significant amount of time to be able to actually plot the drives even over NVMe over Fabric as well as being local on the arrays. Now the one-time hashing process is only done one time. So versus proof-of-work where you're constantly submitting, constantly hashing, proof-of- proof of capacity, you write it once and you read from it

every time a new block is proposed. Now that reading, you get two things. One, find the answer. The deadline is the time it actually takes to read to that spot on the disk. That's part of the consensus. And so just because someone has the right answer doesn't necessarily mean that they would have read to it the fastest, meaning that the location it was on the HDD, SSD, SD card, or whatever storage device that you have, this will allow for an actual true trustless consensus using storage.

Now the power requirements for running proof-of-work versus running proof-of-capacity are starkly different. So proof-of-work requires a GPU, CPU, or an ASIC. Average ASIC use is around 1,700 watts per hour. Now, hard drives' average spend time is about 5 watts. So if you leave a drive on for an hour, it's 5 watts versus 1,700. And so in certain regions, this is where these types of things are good. So BurstCoin, BHD, Boom, and some others are currently using the POC consensus. And then we're going to talk a little bit about BurstCoin and the usage of storage.

So BurstCoin is an open source decentralized blockchain. It literally is everything that Bitcoin has, except it has a smart contract system that's built in, as well as it utilizes a different consensus. Now, when you start talking about usage of a machine, I can still mine Burst and still utilize my devices, as long as I don't need that storage space during that time. So in a lot of cases you will find that and in the current issue which you'll see in a moment when we show some of the people who have rigs for this type of mining, that literally you can mine it from everything from a nest to something like this or to even your mobile phone because you can also do that now with an SD card. Now, this coin came out in 2014.

It was the first to introduce the Turing Complete, as well as introduction into decentralized crowdfunding, as well as cross-chain transactions, which is something that is now being discovered that, hey, maybe all this technology needs a way of being able to communicate. You know, maybe Bitcoin needs to be able to talk to some other chain and vice versa and these types of things are coming because of developers working extensively hard on binary objects and and literally creating a framework for being able to have cross-chain compatibility now uh there's a big difference between a corporate coin or enterprise coin service and these types of chains.

And that's something that we were going to go further in depth with, which is that, you know, if you're only using a blockchain to store transactions or some type of record that's being held internally, you may not even need a blockchain. You may actually just need a database because the blockchain may be a buzzword, but as we break it down, we need to understand the use cases of where these types of things play in and where these types of things play in the storage. Now, going back to the consensus itself, as we'll'll see here this is the process of the actual plot, and I can demonstrate this if somebody wants to see it during the Q&A. But basically this right here is the map to the nonsense, and this is only kind of here as people can download this

from the actual presentation if you want to know more. Unless we really want to get into a topic about cryptographic functions and mathematic probabilities and things like that, I really am going to kind of move forward on that. If people do want to have that discussion, though, I'm open for it for after this. Again, like I said, this is available in the slide show that you can download for this so this right here is the actual further explanation of the mining, plotting

and the flow chart for that so you can take a look as well again also available on the actual presentation so now we're going to talk about some mining and plotting with Western Digital so this right here was a project that we did for another presentation.

But this right here is using one of Western Digital's NVMe over Fabric. Now here you're seeing that we plotted. This is using an Intel Bull Zeon from Supermicro. And we were using all of the NVMe drives over Fabric to plot 7.2 terabytes. It still took us roughly five hours. That is a constant write that's being performed to the drive itself. No, go ahead. It's a back-to-back memory write.

And so what it used right here, you have a Melanoc 900G. This is a Western Digital-related NVMe address, and it's related to a 1..3-square NVMe drive. And what you see here on the axis, this was a GNOME system, right, on the right. Correct. And this is a memory-right performance. So this is a GNOME, which is like computational storage, and that's how the memory-right performs. So it's actually, I was telling Daniel,

it's very difficult. That it is. I mean, the other thing is, versus like with Bitcoin or any other particular consensus, this consensus prevents people from gamifying it because it requires that magic thing that no one can ever get back, time. Because, see, it's easy to throw money and turn hashing power onto something and say, here we go.

This right here actually requires you both to understand that you cannot replot using the same starting knots. You have to understand that you actually have to write to the actual disk. So to do a 51% attack or to actually attack the network requires a lot more than just swinging some hashing power. But we wanted to showcase this. One second. So I also wanted to show some of the mining operations right now. So this is a 7 petabyte miner right here. This is later updated to 16 petabytes that this person is is mining first coin with, as well as VHD. This right here is nothing more than a node droid, which is essentially a small arm,

and then is mining 20 terabytes, as you can see there are the two external Western Digital drives. So from one extreme to the next extreme, people are able to mine these types of cryptocurrencies, utilizing storage. Now, the energy efficiency is where it comes into really play for the transactions. So, versus 1,000 kilowatt hours consumed whenever you're dealing with Bitcoin, versus

24,000 of a kilowatt hour for first. So, when you start talking about on-chain transactions and you start dealing with electrical usage, you know, for enterprise, it's kind of a moot point, right? Like, because you're going to leave your storage equipment on anyway.

You're going to leave your equipment on regardless. So, oh, somebody have a question? One moment. So one of the things that are going on right now as we go back to building the blockchain, using NVMe and OverFabric for the end-to-end solutions, what we're starting to see is that a large number of burst miners are actually starting to switch to NVMe. But also a lot of corporations and businesses are starting to also turn, you know, their cache servers that aren't being used, but like once every week or something that they can't consolidate. Or in some cases, people have been asking us how to actually make plotting faster for them. And you know, one of the things is, is like NVMe allows for them to be much faster in

plotting, but also much faster in their submission time. The other thing is that there are people and businesses who are working on solutions now where they have idle arrays that are being turned into miners when disks aren't being used. Because if they're going to be spinning, you might as well be making money. And so that was one of the arguments that a lot of the industry has been looking at as, you know, for early days of Bitcoin, that was one of the things that a lot of these web companies were doing was literally turning their spare servers, turning them into miners at night because it was a return on the power usage. If you're already, you know, utilizing it, you know, it's about understanding the equipment that you have and understanding the consensus protocols to be a part of them.

Now, when you start looking at NVMe over Fabric, you have a large number of better reads and writes, but also NVMe in general, you know, performance-wise, is much faster than, you know, this little guy here, which is spinning disk. And so the thing about, you know, proof of capacity is that it's able to utilize both, or all three actually, ACD, SSD, as well as MDME. Now here's another drive that we plotted as well.

This right here was only a 7 terabyte, but this was actually a local MDME. This was not of a fabric, and this is only 7 terabytes. But as you can see here, the time increase is shaved off to roughly about an hour, but we're still looking at a 4 hour time in which to plot 7 terabytes. Again, going back, go ahead. If you compare,

so what does the other one know? Yeah. Right? Any of the other ones. Oh. Any of the other ones. Yeah. So if. Oh. Got it. and then we have the local drive. In both cases, here we are using,

and this one was MDM over fabric, with MDM we are using on the Western Digital. And in the Supermicro case, in Supermicro case, we have the MDM Intel, MDM Intel using the local MVME single drive. So in the MVME Republic, it was a real two terabyte drive, four two terabyte,

yeah, four two terabyte drive. So if you look like Daniel pointed out, in a local Intel, one single MVME, you have four hours and three minutes to do the entire memory, And actually, going back to, you know, because I didn't want to really get too in-depth in the cryptographic functions, but you don't want to ever redundantly plot drives. As for you, it just literally is submitting the same that you've already submitted.

So your nonce count, you wouldn't want to resubmit nonce count. And that, again, is something that's discussed further in the actual charts themselves if you want to read over that. Lastly, let's see. Yeah, so there are several types of mining options, just like with Bitcoin. And this is where larger understanding of consensus and larger understanding of what mining actually is. So you have an idea of pool mining, which is everybody participating in a pool are trying to submit as one to get the right answer, but those rewards are shared, right? And the idea of solo or standalone mining, meaning that you are submitting only for yourself

and that you take all. So if you think about it, like winner takes all, right? And that means all rewards, all transaction fees, everything associated to that block is rewarded to the user, not to the pool because you're not being involved with a pool miner. Now, some pools are also using their proceeds and some fees associated with it for different projects and fundraising. So let's go now to talk about decentralized concepts.

Olga, did you want to talk about them? Yeah, so basically the concepts of decentralization are normally based on the three focal points, architectural decentralization, political, and logical. So when you break it down and apply it to MD&E overall, I would say this is a so-called proof of capacity basic concept. So I converted it to what the NVMe brings in. So you have a safe, you have a false tolerance which brings in, you have a tap resistance, our NVMe drive has a checkpoint when you do the so-called trusted computer decoding application. And then solutions, because you have in NVMe spec ability to check through our round-robin algorithm all the collisions for so-called death that is addressed.

On the disadvantages, this goes back to the proof of capacity, too many decisions making and speed of action and duplication. I would say that speed of action and peace of the unique does apply. We do not have a lot of duplications,

but it goes back to proof of capacity. And then decision makers are part of this architectural visualization. If I convert it to MDME, MDME technology or storage technology overall, kept going forward if you look at this, what we discussed right now in the zone and computational storage,

that would bring some of the disadvantages. During this conference this time, All right. So one of the problems is that, let's see, we're going to go back to the performance. What's that? Yeah. Hold on. Sorry, we messed this up. So right now we're starting to see a lot of companies that are starting to take a look at the proof of capacity and mining as well as the use of the consensus and where it stands to introduce itself further into the storage markets. Where NVIDIA and AMD have made mistakes in the past was not acknowledging that the miners were using their products, except for their quarterly reports.

When the price goes up, their sales go way up. And so they made one little line in there. The problem is that NVIDIA and AMD have kept a hands-off approach, and a lot of that came from the fact of the 2009 to 2014 time frame for bitcoin and cryptocurrency as a whole was not looked at as being a shining achievement because during that time only people knew about it was silk road and various things that has since changed from the perspective of individuals but now we're starting to see the rise of corporations we're starting to see the rise of businesses they're're starting to see the rise of businesses. They're entrenched in this technology. And on that note, I kind of want to let Olga finish out anything and then answer Q&A.

I think we did represent the blockchain group of capacity at FlashMemory Summit, so I want to illustrate it. I think we now have formed a group in the blockchain and we've been doing this a lot of new concepts and ideas. And then, we're presenting today, the complements of the Supermicro that we were providing of the latest and greatest in the NBA servers. And of course, with the new digital,

we're just thinking on the new mobile platforms that you see as the real life and what's working today. So I'm going to kind of show the miner again. So as we can see here, the system itself is reading. It's memory reading. Right now it's memory reading. It's getting an error message because I got disconnected off the Internet. Sorry about that. I'll reconnect real quick.

There, something happened. Yeah, so at this point, I'm having a hard time staying connected. I think it disconnected. But basically what's happening right now is we are reading from the actual mining information. We are scanning this drive every four minutes for the answer. And then we are submitting that. We're actually submitting in round times of about 30 seconds and from there we're submitting the deadline as well as

time to read. One second, I'm going to restart the internet. So I wanted to open up to ask anyone if you have any questions, comments, anything that you have that you would like further to explain. How the calculation of the memory, right?

Back to back happening. There is a little video in the slide. Oh, okay. One second. So that actually gives a great idea when you are doing your configurations or anything. Olga, one slide. Okay. I'll go one slide. Yeah.

Yeah. So if you now look at it, if you look at this, this is where it computes the space. So the 7.1 terabytes drive, and it computes your gigabyte storage right here. And once it's finished computing, this is what the hash algorithm for showing, right? The indexing of each of the memory blocks. And once the blocks are all indexed, you are converting it and you start immediately doing the memory write.

So here you'll see the performance going to 2.6 gigs. In that case, 44 gigs, 46 gigs, yeah. So that's the memory write and that's the speed of the drive. And the blocks were 1 gig. And here you know what the memory write is. So you can see that the memory write is going to 2.6 gigs. And the blocks are 1 gig. And the blocks are 1 gig.

And the blocks are 1 gig. And the blocks are 1 gig. And the blocks are 1 gig. And the blocks are 1 gig. And the blocks are 1 gig. And the blocks are 1 gig. And the blocks are 1 gig. And the blocks are 1 gig. And the blocks are 1 gig. And the blocksitter that writes. And it blocks for one gig. And here you will see how that draft is computed.

So that's your index memory, and that's how the memory writes perform back to back. So I find it very interesting when I'm looking for performance benchmarks on the devices. Go ahead. So to do this, you plot the drive first? Yeah, exactly. Correct. So basically you only plot it once. Correct.

After that, you can mine forever. Correct. And then when you're mining, you're doing... Read. Well, it reads the entire plot. So basically you have a choice of one giant plot file or distributed files. Depending on your hardware, you may make a choice that makes sense for you, right? Like, obviously if you're using a Raspberry Pi to mine,

you probably don't want to try to load, you know, 2 terabytes or 10 terabytes into memory. So in that case, you may use may use like two to three gig files um but yeah you basically are reading the entire allocation of hash outputs and then looking for the answer to submit well you're submitting that plus the time it would have taken you to read to that space on your disk so that's called a deadline so just because two drives may have the same answer one may be in a different location on the disk it may take one week one may estimate 35 days so the winner would be the one week deadline time versus the 35 days read time keep in mind even on ac on ACD and SSD, just because something's, you know, we'll say seven terabytes, I may get a deadline for something that would have taken a week to

read to. And that's a legitimate deadline where that actual write occurred on the disk. And so I have to submit both of those, both the answer proposal as well as the deadline to read to that actual spot on the disk. Now, there's no way that, so whether it's SSD or HDD or NVMe, they all participate the same way in the way that the deadlines are calculated based on the scoops. How do you know you're giving an honest answer? You're saying you're submitting the file. Well, we can tell because it will change the key. So when you are plotting, you're using a public key to verify the size of the plot. So if you change the plot upon reading the submission, we know that it had changed.

Does that make sense? Do you understand what I'm saying? So you're using a public key to cryptographically sign your plot size, right? And then if you change or modify your plot size, when the miner actually goes to check it, it would get an unrecognized key signature or an invalid plot size or an invalid plot file because of the fact that it changed its signature. So it wouldn't be valid anymore.

So you wouldn't even be able to submit on it yeah because the key actually is part of the verification process now that's not a private key public key a big difference i want to make sure there's a huge difference in that so if somebody steals your drives for example and they start trying to mine they're still mining for you. They would have to replot all of them. Very key. So is there any questions, comments? Did I answer your question thoroughly enough? Okay, I just wanted to make sure. Does anyone else have anything? I might be lost on the fundamentals.

So in Bitcoin, the miners are trying to be the first to solve a cryptographic problem, the solution of which is independent to each miner. Correct. The solution is generic. When they do, they're rewarded with a block reward. Block reward. And what that does is part of that block includes all the transactions that happened since the last block.

Correct. And I'm not understanding in your system how the rewards and things. It's still the same thing. It's the same thing. It's everything. The only difference is the consensus. And the consensus is like versus proof of work, we're utilizing proof of capacity. Right. And the consensus is like versus proof of work, we're utilizing proof of capacity. And so instead of by writing the Shebol 256 output to the actual drive for the size. So if I have one terabyte like this one right here is plotted out,

or in this case I'll just show you on the file system real quick. One second. All right, so once you've got your drive plotted, right, I assume that you're taking stochastic data that's on the drive and creating a hash across it. Correct. That's the work that's going into. At some point, you have to announce to the world that, hey, I've done this, right?

That you've ordered to plot? I guess. No, you're only announcing when you submit, and you're only submitting, your choice to submit is either to pool or to yourself same as bitcoin so when you're doing any form of trustless consensus and the reason why i have to specify that is there's a lot of them that are trusted meaning like certain coins you have to use their pool and you got to give their devs half of it or whatever anyway um those systems are different so So in this system,

I never have to announce anything to you. I'm only announcing to the network that I am proposing my solution or that I am proposing it to my pool. I never have to announce to everyone. They only see that if I win the block. So this hashing that you do...

When you plot the drive, does that include some transactions? No. It actually is just cryptographic There's one time. No. It actually is just cryptographic output, the function output from the Shabal 256. And from there we take and we read from there. But I think we're running out of time, so I can answer any questions.

Let's drop it off, and then we can stay and do the Q&A. Yeah, if somebody wants to ask questions, I can show you what the actual outputs look like if you'd like to see. So let's do a couple of the final slides. I think we already did them. No. So in the last one. Oh, yeah. Oh, yeah, so the DApps.

So we'll do a quick thing. So the DApps are, you know, obviously what people are going to be using in the future. Also, if you're interested in learning more, you can touch with us. We'll fix that website because apparently it got chopped off. And then, like I said, if you have any questions, we can answer them after this.

And thank you so much for allowing us to present. And hopefully... Thanks for listening. If you have questions about the material presented in this podcast, be sure and join our developers mailing list by sending an email to developers-subscribe at sneha.org.

Here you can ask questions and discuss this topic further with your peers in the storage developer community. For additional information about the Storage Developer Conference, visit www.storagedeveloper.org.