Intel Corporation (INTC) Earnings Call Transcript & Summary

Hello, and welcome. I'm George Crump, Chief Marketing Officer at StorONE. Thank you for joining us today. It's a pretty exciting day for us. We're going to be showing off our new All-Flash Array.next solution. It's going to help you be a better storage hero. If you followed us at all, you know that we talk a lot about being -- making people storage heroes, and we think this will make you even a better storage hero. A key partner in that product is Intel. So I'm really excited to be joined by Joe Nielsen. He is the Business Development Manager at Intel Corporation. Joe, thanks for joining us today.

Yes. Thanks, George. It's great to be with you. We're super excited about what you're going to be talking about. Just quick intro for me. I work in our SSD group at Intel. And the SSD folks are mostly headquartered over here in Folsom, California, north of the Bay Area, although most of us have been working from home for the last few months. But again, great to be here on the webcast. Thanks, George.

You're welcome, and again, thanks for joining us. And then for those who don't know me, I am the Chief Marketing Officer at StorONE. I've been doing that since March. Prior to that, I was founder and lead analyst at an analyst firm called Storage Switzerland. You might recognize that name, hopefully. Prior to that, I was a CTO and a large storage integrator, focused primarily on the storage space. I've been in storage for longer than I care to admit to, but certainly more than 25 years. So it's been a little bit of time. It's funny, we talk about -- you were talking about Folsom. That was the last time I was on a plane, was when we were out there talking to you guys in February. Who would have known now. It's the longest I've been grounded, I think.

Yes. I can't remember the last time I've been on a plane. So yes, that's pretty cool. Yes, I can't wait to get on planes again, actually.

Yes. Getting a little piece and quiet, right? So anyways, let's talk a little bit about what we do here at StorONE. We are software-defined storage consolidation or what we like to call Storage Consolidation Done Right. And if you look at the market, most data centers today have 5 or 6 storage systems, and you're trying to solve specific problems with those solutions. And I think what we're going to show you today is the capability to consolidate that down to a single solution. But what's different about the way we do it is we can start with a single point solution. So you don't have to sweep the entire data center day 1. So if you're having a MySQL issue or an MS-SQL issue or you want to improve VMware performance, we could start there, and then you can expand the use cases all the way up into ML, AI, things like that. And what we really focus on at StorONE is efficiency. So we're constantly trying to deliver the lowest dollar per IOPS available. The other thing that makes us really different is we provide transparent pricing. You go to our website, there's a section there called TRUprice and all the pricing, including the pricing of the new product we're going to talk today is all right there for you to see. We don't hold anything back. We support all the -- I call them the 3 Ps. We support all the various platforms, protocols and ranges of performance. Obviously, here, we're talking about high-performance to even extreme performance categories. We support block, file and object. We support Fibre Channel, iSCSI, SMB, NFS and S3. So pretty much anything you want to do with the system, our software support it. The overall goal is to allow you to simplify operations so that we can just make life easier for storage managers.

Yes. George, I was going to say you guys support the whole alphabet soup.

Yes. Anything with an acronym, basically. We've got it covered, right? So it's totally there. And of course, we support NVMe drives, which we're going to be talking about in just one second. So the product that we've been talking to everybody about is what we call the All-Flash Array.next or AFAn. And it's an Intel-powered, Intel Optane powered flash array. So a lot of Intel logos in here, Joe, Intel Server, Intel Optane, Intel QLC and our software, right? So it's a -- and I think the -- and I'll let you comment on this, too, but from the -- as we started this project and started to really kind of work on it, it seems like at some point, we hit this point where it seems like these 2 products were sort of meant to go together. Remember, that was kind of your reaction as well.

Yes. Yes, we kind of have a mantra in our business group, where we kind of -- our vision of the future of storage is what we like to say, a little bit of Optane with a lot of QLC. And we'll get into what those technologies are, but that's really kind of where we see the future of storage and your platform perfectly supports that vision. So the one thing we really love about StorONE is how well aligned this solution is with our strategic vision for the storage, the future of storage. And not only that, but you guys have some really amazing software that unleashes the performance of the CPU. So when we're looking at performance in your box, the performance, you get basically very close to the spec sheet performance of the SSDs, whereas with a lot of the other storage solutions out there, that are great solutions, you're not going to see the full performance of SSD. So we love you guys for being able to showcase the goodness of SSD.

Yes. Well, thanks a lot. It's funny. Efficiency has rewarded us, being efficient from a software level has rewarded us in different ways and a cost-competitive environment allows us to run on sort of mid-tier Intel servers. And -- but in this world, it rewards us because we don't need a dozen Optane drives to get that high performance. Because like you said, we can hit that spec sheet number or almost that spec sheet number 4 or 5 drives. We're off to the races in many cases. So very fascinating from that aspect. So Joe, let's take a moment here and ask our audience what they are looking for in their environment. So if you could, I would want to know, what is your biggest storage challenge? Is it lowering costs? I would assume with the downturn and the economy in most places that, that could be a big issue, meeting performance demands. It seems like no matter what the economy is doing, we always need storage to go faster. Meeting capacity demand, same thing, right? So we always -- seems like we're always running out of space in the data center.

Right. Rising tide.

[ Running out of ] space -- but yes, exactly. And then data protection always remains the thing. One thing I always remind people about is because everybody's working from home, the chances of malware and things like that occurring or probably improved, right, because the bad guys can work from home, too, right? So it could go either way. And then finally, consolidating storage silos, which, of course, is a big thing for us. So as I look at these early results coming in here, Joe, about 50% of the audience is focused on meeting performance demands, about -- and then split, 25%, 25% between lowering cost and consolidating storage silos. Any initial reaction to those votes?

Yes. It's interesting that it's kind of spread out. Wow, there you go. Now we're at 33% for each of the 3. So yes. And we may touch on multiple points here. So this is interesting.

Yes. It is interesting, the capacity -- and I found this lately that -- and lately, nobody votes for capacity next, but if anything, capacity has become less of an issue because the density of the drives and everything, I think capacity is less of an issue than it used be. It's -- I think the combination of lowering costs and meeting performance demands almost always becomes a big issue. So we'll leave that open for a little bit while we go through the presentation, and we'll wrap it up there at the end. So obviously, Intel Optane, Intel, in general, is a big part of our offering here, Joe. So why don't you talk about -- I've seen you guys kind of talk through this chart before. And one of the things that I see people get confused with. So I think, first, give us a high level description of what Optane is versus a typical NAND SSD.

Yes. Yes, let's do that. Well, and by the way, just to quickly set up for that, I think what we like to -- as you mentioned, a lot of people have seen this pyramid, the hierarchy of memory and storage. And what I like to say is that economically, we all want to have our -- well, we all want to have our data as close to the CPU as economically possible, right? But obviously, it's not feasible to be putting all of our data in DRAM, and it's not -- it's volatile as well. So it's not, not volatile. So therefore, we have SSDs and we have cold storage, lower cost forms of storage. But basically, we came up with this Optane technology to essentially deliver DRAM-like performance, with the nonvolatility of traditional flash storage. So it's a completely new type of media, right? And yes, and I think we're -- I think one thing that -- when you look at the slide, one thing that people always wonder is, all right, they may have heard of Optane technology but they don't realize that there's multiple types of Optane products. So what we're trying to highlight in this pyramid here is that we're filling in the gap between DRAM and SSDs with this new technology called Optane. And there's a couple of different types of products that are filling that gap. On the memory side, we actually have memory DIMMs. They are in a memory DIMM form factor, they plug into the DRAM slots and the media, the silicon in those memory DIMMs is Optane. We call that Intel Data Center Persistent Memory. And then we have traditional SSDs that contain that same media, but they're in U.2, 2.5-inch or Add-in-Card or M.2 form factors, traditional SSD form factors and they work over any PCI Express bus. So we have traditional SSDs that have -- that are based on Optane technology, and we have basically memory products that are based on Optane, and those memory products can be used as persistent memory or in other ways as well or just as DRAM. But we're really focusing in here, for the purposes of this conversation, on the storage version of Optane. So think the Optane products we really want to talk more about here are the SSDs that contain Intel Optane technology. And the one other thing I'll say here is looking at the -- strategically, you look at the pyramid, we -- I was talking about gaps. You've got this gap in between DRAM and traditional flash. You can't make flash fast enough and you can't make DRAM cheap enough, right? And it's volatile. So you've got this gap that we're filling with Optane. You also have a gap between traditional flash storage and hard drives and where -- yes, where we're -- the other big focus we have is filling that gap with QLC-based SSDs, where you're basically -- you're storing more bits per cell and driving the cost down. So that addresses the capacity and cost side of the equation and then Optane addresses the performance side. So a little bit of Optane with a lot of QLC is going to give you that best of both worlds type of solution, which we were talking about earlier.

Yes. Yes. And that's exactly what we really kind of tapped into with this solution is leveraging those 2 technologies together. And they actually really become -- even though they've seen kind of diametrically opposed in a way, they actually complement each other, I think, for those reasons. So I think it works really well. The -- if you look at this, what's been the -- where are you seeing most of the interest? Is it at memory capacity or is it storage?

The storage has been around for a little bit longer. That's a good question. And so yes, there's definitely been a little bit more traction there. The memory side of things is a little -- it's newer. We just started shipping that late last year. And the angle for the memory is that it requires a lot of software customizations to really take full advantage of. You can run it in DRAM mode. So if you're running at DRAM mode, and you can basically use it as lower-cost DRAM. And that makes sense if you're using a whole bunch of DRAM because you do need to mix it with traditional DRAM. So if you -- so if you're running at memory mode, it's pretty straightforward, plug-and-play. You're good -- you're mostly good to go. But there are some platform contingencies. You got -- have to have the right CPU and whatnot. But if you're running it as persistent memory, a lot more complicated. Now on the flip side, the Optane SSDs don't require really any software customization to work. They just -- they'll work like any other PCI Express NVMe, SSD, just plug-and-play, they work in everything. So with that in mind, you can see how the Optane SSDs have gotten a little bit more traction. But again, if you're -- if the solution is really optimized for Optane, you're going to get more value out of that. And that's where I think the StorONE solution really does a good job of doing that.

Perfect. Well, that's a good setup for our next slide. Sorry, I went too fast. So let's talk about the solution. So what we've done here is we've created a solution that delivers over 1 million IOPS from just 3 Optane drives. We actually -- in the shipping systems, we're shipping 4 drives. It delivers a complete enterprise feature set. So with that performance, we're still doing media -- protection from media failure. We're doing snapshots and all those different sort of things that you would expect in an enterprise-class system. We also, again, as I said before, as part of our storage consolidation effort, we also provide all the various use cases. So that's file NFS, SMB, block, iSCSI and Fibre Channel and the object to S3. So it's going to be interesting as we start to have customers deploy an S3 object store on an Optane-accelerated system. It will be curious to see as we see those at scale. And so what we've done is some things in this technology to allow you to start very small. The starting configurations are around 50 terabytes, and then you can grow almost as large as you need to. And so we've done some things here to manage costs and to still deliver full performance and full features. And so that's really the intent behind the product. That's why we call it All-Flash Array.next. It's really, in our opinion, the next logical upgrade to the all-flash array. If you think about it, we've been sort of stuck in the all-flash -- ever since the first all-flash arrays really started to shift, we've just seen marginally faster systems generally driven by more CPU capabilities from Intel, but not so much changes in the technology itself. We've gone down the -- from SLC to MLC, the TLC path, but we haven't really altered the trajectory of these systems, and this is really the first alteration, I think, that we've seen in quite some time.

Yes. It's like all-flash arrays are so 2019.

Yes. Exactly. Good. So Joe, as I said, the Optane is obviously a key part of this. And so I wanted to give you a moment to sort of dive deep on the technology. So can you kind of -- why is it -- so first of all, at a high level -- and correct my math here because I'm doing it literally, I'm doing it off of memory, but the -- if I remember right, it's about 64x write performance on this technology is about 64x faster than the typical NAND. Is that right?

Yes. Yes. The place where these SSDs are really, really going to shine versus traditional NAND, it's your random writes. You can't even compare that. In fact, there's actually multiple specs where even the best traditional NAND-based SSD can't even come close. And it's because these Optane drives are built with a completely different type of silicon. And it's basically -- traditional SSDs are built with floating gate transistors, right? And they're electrostatic. So what that means is they capture electrons and you have basically a charge in a memory cell. And depending on how much charge you have that determines whether you have a 1 or a 0 or a 1, 1 or 0, 1, depending on how many charge levels you can read in that cell. So that's a traditional SSD. These guys are -- the Optane is a completely new material science. And it uses -- it measures the resistivity difference between the 2 materials to determine whether it's a 1 or a 0 and it's a bit addressable. So what all of means is that it's a whole lot faster and you have to -- you don't have to do a lot of the complicated, cumbersome things you need to do with traditional SSDs. So with traditional SSDs, one thing you have to do is you have to erase before you write. And so write speeds are always going to be lower than read speeds, and you don't have to do that with Optane. So write speeds are going to be mostly on par with read speeds. Also, as the slide says this, like I said, a bit addressable, but data is written at a bit level, right? So with Optane, you can address this basically a single bit. With traditional NAND, you write in very large chunks and you read also in big chunks, slightly smaller chunks than writing. And as a result, you have to continually move data around to do garbage collection on the drive, which creates background tasks, right? The other thing you -- the other thing that is another big differentiator between Optane and traditional NAND is that this new media doesn't really wear out. With traditional flash, in order to charge up that memory cell, you have to like push electrons over an insulator on the transistor. And that layer can wear out. Kind of like an old filament light bulb, right, the more you write to it, the more you're going to wear it out. So from an endurance perspective, it's kind of night and day. And what that also means is you don't have wear leveling going on in the background. So there's a number of background tasks that you have to do with traditional flash that you don't have to do with Optane. And you don't need to be reading and writing in parallel before you get to the level of IOPS that you see on the spec sheet. You don't have to build up a really big queue depth with Optane. You can get that performance at very low queue depths, where oftentimes, your system is really operating. So you get that sort of more instantaneous sort of responsive feeling. So fundamentally, that's just -- I'm explaining all these things to just explain why some of the performance characteristics are so different. Like even from a firmware perspective, I believe that lines of code in the firmware for Optane versus a traditional flash SSD is like 30% of the lines of code because there's just not as much that we have to do in the background for these Optane SSDs.

Yes. And I would think that, that's -- I mean, obviously, that's a big thing that we saw in our testing. The interesting thing to me anyway is I'm going to move over to the next slide and I think you guys have got to demonstrate really nicely here. So talk about -- so that's where that 63x is. I know it was 63, 64 performance improvement comes from, right?

Yes. So yes, this is -- this slide is a little intimidating if you're just seeing it for the first time, but I'll super simplify it for you. I just think it's really cool. Like I mentioned, the -- in some ways, no NAND drive can come even close to Optane from some performance characteristics because it's so different. So what we demonstrate here is that we show the responsiveness under load. So you see on the chart this stairstep black line kind of on top of all the other stuff. And what that stairstep black line is showing you is we've got a mixed workload going on. And we increase the write pressure. So we increased the amount of writes that we're doing on the drive from a certain number of megabytes per second to another -- to a higher amount. So as we -- so what we do is we slowly increase the amount of write, the percentage of write happening on the drive to simulate a bunch of different applications trying to write to the drive at the same time. And then we measure -- in the colored lines, we measure how long it takes to get back a read, right? And so -- yes. And so in blue, what you're looking at is, like historic -- that -- let me see -- yes. That was one of our fastest NAND drive in our last generation. And as you see in blue, think of those as little dots, right? Those are -- and that's the measuring the read latency at the time it took to get back to read. And as the write pressure increases, the amount of time to get back a simple read from the SSD goes up proportionally because obviously, reads might get stuck behind writes and then occasionally, you've got to do some of these -- a lot of these background tasks I was talking about in the previous slide. And so not only does the read latency go up, but the variability or quality of service kind of gets a lot worse. Because you don't know when you're going to get those read backs. Occasionally, you have those dropouts where it's going to -- you're stuck in a traffic jam, if you will. Now if you compare that to Optane, you can't even really see it, but the whole x-axis at the bottom...

I was just going to say that we need to point out there for everybody that the red line down the bottom that you can barely see any movement on, that's Optane, right?

That's Optane. Yes. It looks like it's just a -- it's look like it's a, I don't know, like the line for the chart, but it actually is -- those are the dots. Those are -- that is the read latency measurements for Optane. And it's like unphased, right, unphased by all the stuff that's going on. You just get the immediately -- immediate feedback from the SSD. So I think it's just a really cool example of look what Optane can do, right? The responsiveness under load is awesome. And we -- so it makes for a perfect hot tier SSD.

Yes. Exactly correct. And the other thing that's better here, and I don't know if you've mentioned much yet, Joe, is that Optane also has better endurance just life expectancy, right?

Right. Yes, yes. Yes. Like traditional SSDs are typically between -- we expect them in drive writes per day because that -- you can refer to that for a family of drives doesn't -- when you're talking about terabytes written, that's dependent on how big the capacity is. So a family of drives with standard endurance with typically 0.5 drive write per day to 1 drive write per day. And then the higher endurance drives are maybe 3 drive writes per day, we used to have a 10 drive write per day drive, but we just shift to 3 now and then you can overprovision, you can dial it up to 10 if you need to. But these guys are at 60. And around 60 drive writes per day for Optane. And I can tell you that we definitely have the capability to make it a lot higher in future generations. The underlying media itself is extremely endurant. It's very, very difficult to wear these things out. So again, you're talking about the how random writes are really where Optane shines. And writes -- the way you're going to wear out a drive is through random writes. And so that's where you put an Optane upfront, right? And you use that as kind of your hot tier, let it absorb, writes all day long, and you're just really not going to wear it out. It really shines in those types of workloads.

Yes. Perfect. Well, that's a good setup for kind of how we did it, right? So the first thing is we use Optane as storage, not as a cache. And so that's an important differentiator that I think a lot of times, it's really easy to kind of skip or miss. In most cases, what people will do is they'll use the technology, a lot of these technologies but the Optane, in particular, as a cache. And the problem with using -- I mean, caches are great. And like in a single-server environment, I know that you guys have a great solution there, but the challenge is that there's no media protection typically. Typically, it's one drive and you're doing something. And so you've got two choices. You've either got to get data out of that cache as quickly as possible, just in case there is a failure or you've got to do a write-through, right? And so basically, you're going to write it to Optane and then through to the next layer down, or which in most cases, would be a TLC-based SSD technology. So what that means, though, is at that point, your acknowledgment is coming from the TLC that has all those write cycles that we were just talking about. And you lose, I would say, most, if not all, of that huge gain in write performance as a result of that. So because we use it as a tier, now to be fair, we have a greater responsibility. I've got to protect that tier, right? It's storage, not cache. So that means I have to have media redundancy in it. And that's why we ship the units with 4 drives, 3 terabytes of capacity there. And 1 of them typically is allocated to media redundancy, so we can survive a drive failure and things like that. We have full data protection there. So the data can stay on that upper tier for as long as you want -- as long as it can, I should say. And the real goal is to -- if you look at the life cycle of most data, it's typically created very, very active during the initial phases at rate after creation, modified a lot. And then typically settles and then in many cases, never changed again. And in some cases, never even accessed again. So what we want to do is give that tier enough -- to be big enough that, that can occur, and then you're ready to be able to move it to a lower tier. And so this is a key advantage because in our software, we're taking full advantage of the full write performance of the system. And as I have written there, and we talked about the 1 million IOPS. The 300,000 sustained write IOPS that we commonly see with the system is -- now that's the 4 drive system. But that's -- I don't -- Joe, I don't even know -- I don't think I've ever seen a NAND vendor even publish a straight write IOPS test. I mean it's almost everything I ever see, it's mixed because you want to get -- you want to give the flash [ dying ] like a fighting chance to catch up, right?

Right. So you don't want to expose the weakness of the flash.

Yes, exactly. So in this test, and I'll show you the benchmarks here in a second. In this test, this was full-on nonstop writes, just going go and going. It's just -- it was crazy. And so to deliver that kind of performance is simply anything in the hundreds is just phenomenal with this type of technology.

Yes. I would say that's heroic to get back to your theme.

Yes, exactly very heroic. Thank you. You might have the new CMO job here. So the big question is, how do we do it, right? And so we have this QLC smart tiering technology, and we call it S1:Tier. And so we've done some very specific things as it relates to the combination of Optane with QLC, right? And so if you think about QLC as a technology, it's very high density, it's relatively -- it's certainly low-cost in the NAND family. But it does have a much more limiting write capability. So we've got to manage it a little bit different. However, the other advantage with QLC is its read performance is very -- is good. It's not going to be an Optane drive to drive, but it's not bad. So the -- so what we try to do is leverage all of that. We came up with a 2-tier mechanism that works quite well. So what happens is we have a high and low watermark. You write data to the Optane tier up until the point that you fill the watermark up. Now in -- I've had people ask me how long will that take, the answer is it depends, right? In some environments, it might take a week. Other environments might take a day. You just don't know. It might -- in very, very active environments, it could take an hour. It just -- again, it just depends on the environment. But in most cases, it's probably closer to a day to multiple days. And then once -- and again, remember, it's fully protected on that tier. We've got drive redundancy in place. Once we hit that high watermark, we then move that whole chunk from the high to the low to QLC, and we write a nice, big sequential write and pop quiz, Joe, what's the best way to write to a QLC drive?

In big chunks. And sequentially. Yes. Yes. You don't want to be doing random writes on a QLC because -- and especially with our 16-terabyte QLC drives, yes, it's sequential is the way you want to do it.

Yes. So that's what we do. We write this nice, big sequential chunk and straight down to the QLCs. That does a couple of things. Number one, not only is it the optimal way to write to QLC, but it also sets us up for very good read performance, right? Because one of the things when you start talking about hybrid systems, of course, is, okay, what's my performance going to look like when I don't get the access data off of that upper tier? It's a very fair question. It frankly is probably what hurt the flash and hard disk hybrid systems back over 12 years ago because the hard drive performance was so bad. So in this configuration, the performance off of the QLC tier because, a, there's tends to be more drives; b, I've written this nice beautiful sequential write; and c, our software is kind of good at getting really good read performance. We can deliver the data off of the QLC drive at nearly the same speed as we can off of the Optane drive. In fact, if there are more QLC drives, which would be common, we'll actually kind of deliver the same performance. And so what -- that fundamentally changes everything we've ever thought of about a hybrid system. I now can serve data directly off the QLC tier. I don't need to worry about promoting it up to the upper tier then moving it up and then having to worry about it reaging and going back down. I just serve it directly to the QLC tier. Since most accesses to the lower tiers are [ referential ] in nature, over 80%. What that means is when I go to -- all I'm really doing when I'm accessing data off of that tier is I'm kind of looking something up, right? I'm not actually going to change anything. So why go through a whole another write cycle if I don't need to? So that's the first benefit we get there. We also get a lower latency because I'm not moving data all over the place. The third benefit is, I don't -- I now have a much higher efficiency on my upper tier, which again is Optane, because I now -- most hybrid systems have to pay attention to access date. Because if you're accessing a file a lot, and your lower-tier is slow, like say, a hard disk drive, you have to make sure that, that data isn't on that lower-tier as well. Well, since my lower-tier is almost as fast as my upper tier, I don't really have to pay attention to access date, I just have to pay attention to modification date. So if you look at the percentage of data in the data center that is unchanged, it's typically about 80%. If you look at the data in the data set, so about 20% of the data is active at some level. But if you look at the data that is presently active, meaning being created and presently being modified, it's generally well less than 5%. And so what that means is now I have this incredibly efficient upper tier that's fully protected and then leverages the QLC tier. The net result of all of this is I can deliver a solution -- this solution that does 1 million IOPS, 300,000 write IOPS at a price less than all-flash arrays.

Yes. And George...

Can you comment on that?

Yes, if I may. To me, it's just way smarter than building an all-flash array with all TLC SSDs. I mean because you have so many trade-offs there. You're using traditional NAND that will never be able to do some of the things Optane can do. And yes, Optane is a little bit more expensive. But proportionately, the number of gigabytes versus the QLC gigabyte is low. So you had a slightly higher cost media with the slightly lower cost media, which kind of equals out a little bit. It comes -- it's just a way smarter way to do it versus all TLC for a very similar type of price point.

Yes, exactly. And I think the key -- if you look at historically in technology, the next-generation has always allowed customers to -- even though it's faster and maybe a little bit more expensive, it always allowed customers to lower pricing, right? When all-flash arrays came out, I had to go through and explain to customers who, 3 years prior to that, I had implemented a 300 hard disk-based system and short formatted all the hard drives, so I could get as much performance as I could. And then I had to explain to them that, hey, guess what? With us, even though per drive it might be a little bit more expensive, I don't need as many drives. And we're kind of hitting the same thing again. Now the key difference, though, I think the key enabler, maybe is a better way to say it, is that you have to be able to deliver close to what the drive can do, right? If you can only deliver 20% of the drive's performance, it makes it much, much harder to justify. The fact that we're getting these numbers with 4 drives, that really helps that equation make a lot more sense faster.

Yes. You can't have really bloated software on top of these things.

Yes. Yes, you guys eliminate all the room to hide.

Yes. Right.

So let's talk about the results real quick. But before I do that, I want to talk about the protection because we've hinted that a little bit. So again, this is enterprise-class stuff. We call it our Five Star Safety Rating. We've got 5 key features, obviously, there. We've got full HA. So if you can buy this as a single node or dual node and we've got full HA there, we've got RAID, what we call vRAID, which is an advanced form of Erasure Coding. So I can deliver full data protection. One of the nice things about that technology is that we have the ability to mix media sizes. So Joe alluded to the fact that there might be higher capacity Optane drives in the future. Well, if you start off with, say, 4 750-gigabyte drives, which is what we use today, and in a year or whenever -- I'm not going to commit you to a date, Joe, so I won't get you in trouble. But in a year or something, you guys come out with a higher capacity drive. With this technology, we'll be able to leverage the full capacity of that drive not have to format it down to match the older capacity drives. So you get the full benefit. We have unlimited snapshots, so you can take a snapshot every 3 minutes. You can keep that snapshot. And definitely, we have customers with hundreds of thousands of snapshots. We have a couple of systems in our lab with over 1 million snapshots. We have replication, which is the ability to replicate from anything to anything. So that's interesting in this use case because in your DR site, you might not need that -- you might just want to be hitting that DR checkbox is what I call it. So we can replicate from anything to anything. So it gives you the flexibility there. I would say you probably want -- what you don't want to do is have an All-Flash Array.next in your primary data center and replicate to like a hard disk-based system because then if you actually do have a disaster, people will throw things at you. So you want to have a little bit of Optane in there, I think, just to make sure that you can get through. And what I say for last is really important, and it frankly, is something that Optane -- this is another situation where the synergy between the 2 products was remarkable. We have never used a RAM cache to enhance performance. We did excellent performance to standard NAND-based flash without needing a RAM cache, and that gives us a lot of benefit as far as data integrity and things like that. When we acknowledge a write, we're actually acknowledging a write. Where it really helps is when we're talking about something like Optane because the technology is so fast, the loop to RAM to cache it actually ends up costing your performance because you can't build the RAM cache big enough plus you're going to have data protection issues. So just writing and taking full advantage of that, again, that 63x write speed is a huge capability for us. So I think that is another area where the 2 products almost looked like they came out of the same lab, like 2 people were thinking about the same thing at the same time.

Right. Nice.

So the -- so -- and then -- so let's talk about numbers. It's always good to talk about numbers. So I'm going to start at the bottom. So the random read performance was over 1 million IOPS. 4K reads, you got all the datas right there. And then you can see up toward the top, we have 2 different write numbers of 320,000, 310,000 IOPS. And so that gives you your range. Now what I want to point out is 2 things, is there's these numbers in the middle there you might not have seen before. I know I didn't see it, called the RE-Write number. And when I saw that, all I did and maybe you do this sometimes, Joe, I said look, I need the performance test to test this, this and this. And one of my business was I got to be able to show what happens when we tier, right? And -- but unfortunately, because I'm not a detailed oriented guy sometimes, I didn't explain exactly what I wanted there. And so what the gentleman did that was testing this was team filled up the Optane Tier and then let tiering happen to the QLC tier, right? And in a normal data center, you'd either have enough pauses in the action or there would be enough read performance that, that evacuation to the lower-tier would literally be unnoticed, right? In this scenario, you kept writing to the Optane tier nonstop. And so what happens is we got into a state where on a per second basis, we're filling up the Optane tier forcing an evacuation, writing to QLC. Remember, we also have data protection and snapshots all happening at the same time. So we're also doing parity calculations on the upper tier, doing dual parity calculations on the lower tier and all of this is happening. So you've got like for every write, you've got 5 other writes happening because of the state of the system. What's incredible to me is that in both those tests, we still maintained 155,000 and 220,000 IOPS, right? I mean that's just as a testament to the technology from a hardware perspective and the software. I think the thing that really jumps out at you here, if you don't -- you just got to make sure you don't miss it, is look at the latency column. And all of that activity, when we're doing all of that work, the latency is remarkably consistent. Right? And so that's -- I think for most environments where you're going to see the benefit of this solution is -- we talked a lot about IOPS because people -- it's something you can grab on to and hold. But I think where you're going to really see us is the low latency that it brings. That's where you're going to see applications really begin to scale. Any thoughts on these numbers, Joe?

Yes. I just think it's a pretty powerful example. I don't think you'd be able to do what you just described with any kind of -- with the traditional NAND media. I think this is what the sort of capability. It really -- this is really what Optane was built for. This particular tiering example is a perfect one. I love it.

I think NAND would have melted, actually at some point. So I think they ran that test for like 48 straight hours. So it was kind of funny. I felt sorry for it after a while. So the intent is really great performance here, really excellent, consistent latency, which is really what can get you in probably the most trouble. So where does this make sense, right? I mean we talked a lot about the technology. It's less expensive than an all-flash array. So you could do a literally, a box-for-box trade out and benefit from it. So where we're seeing the most interest is right in the middle there, that the ability to scale VMware, Oracle, Microsoft SQL, MySQL further. So lower the licensing cost. So again, this is less expensive than your typical all-flash array, and it can scale those environments much further so you can lower the licensing costs. We have 1 customer using -- that's looking to use a product that calculates that they'll be able to eliminate 2 physical servers and up to 16 cores of licensing because most of these products that I have listed there tend to license on a per core basis. Well, the savings and software is going to allow him to buy the system, right? So to some extent, you'd almost say he's going to be getting it for free. And then when you're ready, you can expand into analytics, AI, machine learning type of workloads because again, we're talking about performance that can sustain that kind of activity. The third bullets in there, just -- obviously, we live in a different era today. Joe and I were joking about not getting on planes. That also means my engineers can't get on planes. We can't send a team of people anywhere anymore to do this sort of stuff. So the system is designed to arrive in your facility. All you've got to do is give us power and network access and we can finish the installation remotely. We can support it remotely and then also that means you can operate it remotely. So you really can't race to the data center every 15 minutes anymore, so you need a system that you can manage and maintain remotely. That was really part of the design here and why the -- I think the commonality of components and things like that is so important to the system. So anyway, so that's how that goes. Any comments, Joe?

Yes. Yes. No, I think we've mostly covered it. Yes, pretty exciting. There's multiple flavors of the Optane drives, different configurations, but right now, the Optane drives themselves kind of range from 375 gig to 1.5 terabytes. So that's kind of where we're at today with the Optane side of things. The QLC range is obviously much bigger.

Yes. And we're ready for the next big ones to come. So hey, let's wrap up on our poll real quick. It's kind of interesting. We almost ended up with a 3-way tie. For lowering cost, meeting performance demands, improving data protection and then consolidating storage silos. So there's a couple of data protection guys snuck their way in there. And then the -- so that's -- so I want to thank everybody for voting. I'm going to close that poll out now. Joe, you're up for a few questions?

Yes, let's do it.

Okay. Let's see. Okay. Let's -- the first one looks like it's for you, Joe. How long will the current generation of Optane SSDs be shipping?

Yes. So that's a good one. We're actually -- we're going to -- we're shipping these current generation SSDs much for a little bit longer than we traditionally do. So the P4800X which is what we call our current generation Optane SSDs, those will be available through all -- we've been shipping it for a while now, that family for a while. And we will -- we're committed to shipping those through all of next year as well. So I don't think we're going to be EOL-ing those products until 2022, maybe early '22, not sure. But you can -- like, if that was something you're buying now, you could say if we continue to buy that, we are working on our next generation, and there definitely will be an overlap. The Optane drives, since they're so new, we're not refreshing them quite as often as the more traditional flash drive. And so it kind of turns out that you get a little bit of longer life with those particular products, which I think is convenient for customers. And we're kind of enabling a longer overlap between generations.

Yes. I think that's good because one of the things we hear a lot -- customers are concerned about, what I would call like future-proofing and things like that, so the longer life of these systems. One thing I didn't mention that I probably should have is in that testing, all the CPU and memory utilization was on the storage server, let's call it, was very, very low. It was below 30% on both memory and storage. So we have plenty of room for growth. We can add a lot more capacity systems and not worry about tapping out those CPUs. So that's also a really good thing. Next one is going to be for me. It looks like how have data deduplication and compression handled? Great question. So as I said at the beginning, we are very focused on efficiency. So what that means is we want to put efficient algorithms in our system and things like that. So we just don't see a lot of value in broad spectrum data deduplication, in particular, primarily because the cost of these systems now, when a NAND was $14 a gigabyte or more, anything you could do to lower the cost made sense. Now that we're $0.30, $0.20 a gigabyte, I don't think it makes sense anymore and doing something different, like moving data to -- from Optane to QLC, we think, makes more sense. You also get into an overhead issue as these drives like Optane get faster and faster, running dedup, just don't have the -- like I said, you don't have latency to hide behind to do those things. So we'll probably eventually bring it out, but it will be something similar to what we did with Erasure Coding, where we'll work very hard on the algorithm to make it much, much more efficient. I will say, though, that our snapshots are unlimited and nestable. So probably the one big area where you're going to see a big gain in efficiency is like a VMware or a Hyper-V environment. Well, with our software, you could build your first Windows VM, make 1,000 snapshots of it, which has no cost from a capacity perspective and then start customizing each one of those VMs as you see fit. So very, very -- and maintain all the performance that we're talking about. So in most cases, it becomes a nonissue for our customers. Let's see, Joe, I think the next one is going to be for you. What form factors are shipping for Optane SSDs?

Yes. Okay. So yes, for Optane SSDs, there are 3 different form factors shipping. Most of our volume for -- and in fact, for all of our SSDs, not just Optane, most of our volume is on your traditional 2.5-inch U.2 PCI Express SSD 15 millimeters thick, just the regular old rectangular box, right? We've got that form factor for Optane. That's what we sell the most of, but we also have an Add-in-Card, PCI Express [ by 4 ] Add-in-Card, kind of half-size Add-in-Card. We actually have kind of been moving away from building Add-in-Cards, so you don't really see that form factor for traditional SSDs much anymore. This is one of the last -- this is the last Add-in-Card that we're selling. But you got [ F2 ] Add-in-Card. And we do have an M.2 version of Optane as well. It's the 110-millimeter version of M.2 SSD. But U.2 is the way to go. I forgot to mention, we do also have 100-gig as well. That's -- we go as low as 100-gig for our data center Optane SSDs.

Perfect. Thank you, Joe. So the...

By the way, I'll...

Go ahead.

I'm sorry, George. Yes, one thing maybe to point out is we also -- in addition to having that Optane memory and Optane SSD products for data center, there are also Optane client products, Optane products for your notebook and your desktop. So I just wanted to point that out, just in case people don't get confused. If you go online and Google Optane, you may come up with like an 80-millimeter M.2 Optane NAND hybrid product, which is something that is targeted for your notebook and not for the data center. So I just wanted to point that out. There are some other form factors and other Optane types of products that we haven't talked about here because they're not -- they're built for your client notebook and not for a data center.

Yes. I can't wait to get a notebook with Optane in it. I'll be first in line.

Yes. Yes.

It probably boots in like 0.5 second or something. Let's see the -- so the next question, I'll take this one. How would I back up data on the system? It's so reliable, you never need to back it up. No, that's not true. It's -- so again, we have the unlimited snapshots. You can use those snapshots essentially as your day-to-day protection. We get -- I think, as I said earlier, we could take a snapshot every 3 minutes. Those snapshots are rewritable. And you could take a snapshot and move it to a backup software. So if you're using Beam or Commvault or something like that, you can just back that snapshot up. So you have an independent copy of it. What we tend to think, and there's a paper. I didn't mentioned the attachment. So there was a whole attachment section with tons of information to help yourselves out there. One of the webinars that are in there is a webinar we did on how better primary storage can reduce backup costs. And we talked about this specific issue. So again, we're not saying eliminate back up but we can lower the cost and lower your dependency on backup because of the levels. I showed you that Five Star Safety Rating. That's a good example of that right there. That one. Joe, next one is for you. What CPUs will Optane SSDs work with?

Yes. This will be a quick one. The answer is any CPU, any CPU because as long as your system supports PCI Express and you have an NVMe storage driver, you can work -- you can support an Optane SSD. But I think the reason why some people tend to ask this question is because the other Optane products we talked about, the Optane memory product, the Optane -- we call it Optane Data Center Persistent Memory, the one that looks like a memory DIMM, those -- that product doesn't work on a PCI Express bus. It connects to the CPU over the memory bus. And if you don't have a CPU with the right memory controller, then you won't be able to use that particular product. So there are CPU contingencies or dependencies for the Optane memory but not for Optane SSDs.

Great. Okay. Let's see. So the next question, I'll take this one. How does your replication work? It's available in 2 flavors. We can do asynchronous or synchronous replication. And soon, like in less than 30 days, we'll be able to do cascading replication. So all of those capabilities exist. If you're a follower of StorONE, you know that we do a software update on a quarterly basis. We don't wait a whole year to do stuff. So we're putting new features in the product quite quickly. And so that cascading replication is next on the list. But today, we have synchronous and asynchronous, so you can do a distance-based type of replication for DR and synchronous for protection from an internal failure within the data center. So that's there as well. Joe, we got like 2 minutes left. Why don't you take us out with what are the endurance levels of Optane SSDs.

Okay. Cool. Yes, endurance is another area where Optane shines. Like when you're getting hit with all those random writes, that's what you're going to -- that's what's going to burn out a traditional flash-based SSD. But with Optane, the current drives are spec at 60 drive writes per day, so multiply that by the capacity of the drive, multiplied by the number of days in a 5-year period, which is the warranty period, and then you've got your terabytes written for the drive. So basically, 60 drive writes per day is pretty good, pretty solid. Not -- I don't know of any NAND drive that can really come close to that. So that's highly, highly endurant. And I don't want to make any announcements, but I think you'll see a good chance that number will be going bigger with -- in the future at some point.

Okay. So thank you, Joe. So just some quick wrap-up information here. I put all our contact information on the screen. There's -- like I said, there's a deep list of additional information that you can click on. There's PDFs and download links to go to. There's -- our pricing site is in there. So you can go price one of these systems out by yourself. It's called -- the link is called TRUprice configuration, take you right to it. You can see how the pricing compares, and then lots of other information. There's ability to reach out to us through those lengths as well. And then there's different ways to get us on Twitter and LinkedIn. And then Joseph was -- or Joe was bold enough to put his own e-mail address on the slides.

Yes. Any other questions, feel free. Yes. Feel free to shoot off your questions my way.

And speaking of questions, we didn't get to everyone, so we'll take those questions. We'll get to you off-line. And so we'll get answers to your questions. So Joe, first of all, thanks very much for joining us on the webinar, and of course, for all your support with this product launch.

Yes. My pleasure. It was great joining in, George. Thanks a lot.

No problem at all. And I want to thank all of you for attending today. If you need more information, there's a contact information on the screen. For now though, have a great day.