 Next up, we have our next guest is Scott Shanley, the director of technical marketing, S-T-E-C. Hey, come on in. So I brought show toys just in case. Ooh, I like show toys. Hard drives. Oh, okay. This is Scott Shanley. You're the director technical marketing, right? It's Scott Shadley. Oh, no worries. Sorry. Everybody pronounces it wrong, so it's not a problem. Just slap me. Not on camera, though. That'll probably go a little upset. Yeah, it'll be a big show. It's interesting. My last name and the company name both have the same thing in common. Most times, nobody knows how to say you don't want to. Oh, no. Is it STEC or S-T-E-C? There you go. There's the question. Yeah, exactly. STEC, S-T-E-C, S-TEC. Officially, it's S-T-E-C, which stands for nothing but our stock ticker, so. I got it right. So it's always entertaining that way. So describe your job a little bit and kind of go into detail. So my job, I'm a director of technical marketing. My job is to go out to the world and educate the customer base on what a solid-state drive is and what S-T-E-C does differently than the other companies in the marketplace today, so. Do you feel still fine that people don't know what solid-state does? It's interesting. There's quite a few people in the industry still that don't know what a solid-state drive is, let alone what it does. So that's why, for example, I brought in a show twice, so I could say, here's what an actual solid-state drive is. Nowadays, we're getting to the point, finally, this year, probably into next year, where it'll be more of, well, which solid-state drive do I want? Why is yours better? What are you doing to make something more unique than the other people? That's really where the innovation of the solid-state technology is going today, so. So you guys are a lot younger than I am, but when I first started in this business, these things were like refrigerators, so I was filled with solid-state and they plugged into a mainframe and they were really, really expensive, really fast, but really expensive. But the big thing is they weren't non-volatile. I mean, they kind of were. You had this big battery backup that you could put in, but they weren't persistent. Exactly. And that's a big change, isn't it? Yeah, and it's interesting because non-volatile memory is in the way that it started out as a NOR-based product and it went to a NAND-based product. They're different internal architectures of the media, but it allowed for the growth of the gigabytes while keeping the size fairly small, so. But it's also a misnomer that it's truly non-volatile because there are wear-out characteristics of SSDs and that's what everybody gets really worried about in the industry. Oh, I hear they're going to wear out too quick and so those are things that we try to do our best as far as the education front goes today. I tell people they do have a finite life to them, but we have ways to predict it, ways to make it manageable, and we've even got technology now where these things are rock solid for five years, so we've eliminated those concerns for a lot of our customers. Yeah, so go ahead. Go ahead. Yes, exactly. Well, I was just going to say before we get too far into it, why don't you go ahead and do a little demo and describe and use the toys. Use the toys. So for most people in the marketplace today, I don't know if I can see where they're at. This is what we'd call a traditional two and a half inch hard drive form factor. Normally you'd see it anywhere from a 9.5 millimeter. Hold it right up there and Mark will catch you. 9.5 millimeter drive versus a 15 millimeter drive, SAS versus SATA, those kinds of things. There's different interfaces involved, but we did these little guys to kind of show through of what the, there we go. See how the complexity of building these things are. There's multiple PCB layers, there's controllers, there's memory, there's flash. One thing that we do uniquely is we build everything internally. So I don't know if you can get a good shot on this one, but yeah, this, this way. This guy right here is actually our controller device. We manufactured ourselves and that's what sets our parts up different from other SSDs in the markets. We design and manufacture our own controller, the PCB and all of the hardware that goes with it, the only part of the SSD we don't physically own is the NAND devices. And we have specific partnerships with our customers to status. So what does that do for you guys other than help your margins be better? What does it do for your customers? What's the value in you designing your own controller? So the controller is really the key to the solid state technology. It's like the controller is the key to the compelling architecture, the ecological architecture. How you manage the media is really what's important. It's the brains. It's the brains. And so by doing the brains ourselves we can add layers of complexity or sophistication that help benefit the customers and be able to give them these long term lifetimes. Another unique device we have out, I don't know, you can see this one. This is a huge beast. This is a traditional three and a half inch drive. So you can see the progression from three and a half to two and a half. And now we've even got a little guy the 1.8 inch in comparison. But this guy is actually made of DRAM. So most people say, well, why would you put DRAM in a solid state drive? Because DRAM is not non-volatile. We've designed this one so that it runs on DRAM during the operation of the device. But when power goes away, whether planned or unplanned, it backs itself up to flash. So it becomes a non-volatile DRAM cache. And it works really good for metadata. What's the ratio of this? Is it one to one? It's really one to one, yeah. So this, because of the density differences between DRAM and flash, even though it's this big, it's still only holding gigabytes. But it gives you a lot of ability to get really fast access data and absolutely nowhere out. So what are you learning? When you guys, I actually wrote a piece when you guys sort of announced your first OEM. I said, you know, I landed a haymaker. All right, because it surprised everybody. Yeah. And it's like, whoa, you know, solid state's back. And so what have you learned since then? Because everybody said, oh, disk is dead and spinning disk is doomed and your stock price went crazy and people calmed down a little bit. Yeah. So take us through sort of what your learnings and your initial launch. So interesting enough, I joined the company right about the time that that major launch occurred. So I was actually, You were excited. Well, that was fun, yeah. I sat through the last four years of it. But really the idea is people need data faster. The data transformation they've been talking about here at the show for the last couple of days. Flash Media gives you an opportunity to get really fast access to information, but it's not gonna keep up with the capacity rates of three terabyte drives. It's not gonna keep up with the cost points. There's never gonna be a true crossover of flash to spinning disk. So it really becomes, you have to figure out the best way to do the hybrid architectures and make sure that flash mirrors and maps in well with the spinning media. We've never had a play in our market portfolio that we're gonna say we're kicking spinning disk completely out of the market. It's always been, here's how we're gonna help augment, make things better. And you know, that whole concept of the tier zero, tier one, whatever you wanna call it, that the SSD fits into. And we make it in every interface, every capacity you can imagine for that purpose because everybody's architectures are unique and different. So you say that and you're not gonna kick in. And I understand that's not your strategic plan, but isn't that inevitable in certain segments is like, you know, the iPod? Yeah. Right? You used to have a disk in it. I still have one. I do have a disk in it. So slow to get going at the shake at some times, which is not a good thing to do with disk. Yeah, I find it a little surprising that you said that. Well, so there's different aspects of the model, of the space, right? So the consumer space. Sure. And there's an even an article posted earlier today about disk is dead in the consumer market spot. Well, for the iPod, the iPad, for even desktop computers, it's very, very viable. Because you can get enough memory locally into the system with the flash-based media and get it at a cheap enough price point now with MLC media to make it viable. But in the grand scheme of the large storage arrays where we're getting Zeta bytes and Peta bytes and Geopytes or whatever you put up on Geopytes, that he put up on the screen there. That's a new one. You still have to have something that's a larger capacity device to make that really work. And so solid state drive will always be there. It will always increment up with it. It'll grow in the amount of it that you have in a system. But it's never going to be a 100% replacement for it. There are systems that can utilize it that way. And there are industry markets that need to do nothing but solid state like that. The heavy duty stuff, the industrial stuff. At a high level, if you look at the big bucket that is spinning disks today, I mean, the percentage of that big bucket that is flash or SSD is tiny. It is. Today it's very small. Less than 1%, is that true? And if you look at all, yeah, pretty much. If you look at the overall scheme of it, but even, you know, you look at any of the industry analysts, they're all saying, well, we've got this great little hockey stick about to come up. So everybody, one of the first questions we usually get asked by a lot of the analyst market is, well, you've got a lot of competition on your shirt tails. And it's like, well, number one, it's nice to have them with the target on my back. I'm not shooting at someone else. Competition is good. Competition is good. And having the competition that's viable, having two sources in a spot means there's more bites for everybody to get at. So we've done the numbers, we've looked at the market, and yes, there's a continued huge spike to take place, bring in some other guys to help bolster the market, give everybody more confidence that we are knowing what we are doing with the technology, and we'll continue to grow together. So do you think wherever that point is not of, let's say it's 1%, just for the sake of argument, do you see that growing, that percentage of the total bit bucket, or is it more of play, it's going to stay 1%, but the whole market's going to grow because of the zettabytes and the better it goes beyond that. I think it really is going to be a little of both. So you're going to have, the continued market growth is going to be huge, but the amount of SSD that you're putting in the system is going to continue to grow. So slowly creeping away. Yeah, whether it be by sheer capacity versus number of drives, or whether it be by the shrinking technology, the geometries of the flash, we shipped today in the Dell Compellent System and 146 gigabytes. We're working with them on their next platform right now. It's a larger capacity. So the bit go up, even though maybe the drive count doesn't, the physical capacity's still increasing in the system. And are you able to increase the capacities at the same rate as spinning disk? Are you a little bit behind, a little bit ahead? So if you look at it, even four years ago, we were talking four, eight, 16-gigabyte drives, right? You know, you're talking to huge monsters. But even in this form factor, we're getting them at 40 gigabytes. We're up to 800 gigabytes today of user space capacity in the drive. And three terabyte spinning disks just came out. So we've made a very large jump to catch up to it. I don't know that we're going to see that, again, that capacity pairing of the devices, but because the geometry shrink and the technologies improve, we can get much closer to high capacity drives very quickly. So as an example of encroaching, why would anybody, you know, in the next, let's say two, three years out buy a high spin speed fiber channel device that's, you know, I don't know what they are today. I think 300 is the max today. Why would anybody buy one of these? I mean, at some point, your price is going to get down there. It doesn't need, does it have to cross over? In other words, does it have to be cheaper? No, actually, if it just gets close enough, I don't know what that is, 10, 15%, people say, I'll take that. Because there are benefits to it, so they absolutely would. The same decisions we're making with laptops. So I should like the MacBook Air. It's a little bit more expensive. I don't get as much capacity. Okay, I'll go for it. Yeah, exactly. I bought the Air over the Pro myself, so. But, yeah, exactly. It's nice, you have to look at it. Yeah, exactly, this is true. But you have to look at the concepts also of where you're going with the market space. So yes, you could get closer on capacity, you're going to get dollar per gigabyte equivalency. Again, you're not likely ever going to get dollar per gigabyte equivalent, but look at the overall picture, the TCO model, the ROI, whatever you want to call it. The value. The value of it, right? So over here on the Solutions Expo, we've got a demo setup, one drive against 24 HDDs. I get just as much capacity out of it because they're all short stroke to give them their performance play, just like they do everywhere else in all the systems. And it's faster. It's 5X faster. So if you put one versus 24 and you do a couple of those, you know, you get to very easily come to the point where the dollar value makes perfect sense. The dollar per IOP, the dollar per watt, or even like OLTP transactions, the dollar per transaction goes from dollars per transaction to pennies per transaction. So we've got a whole bunch of success stories with many different people. One of them in a compelent hometown of Minnesota, the Wild, the Accelerina there, put just two drives into their existing compelent architecture when the SSDs first launched and made it so that their concession stands, the slowest part of the concession stand was now the beer tap, not the cash registers. Because their whole back-end architecture was able to be improved that much just by adding a couple of SSDs. So that's the real value out of these products is we love to displace the entire array. I mean, you want to give me a purchase order for 7,000 drives, I'll take it. Just three or four drives just to give the customer what they really need and making sure that we give them the right amount. So you don't want to oversell it, you don't need to undersell it. I want to go back a second. You talked about how, you know, with the SSDs, you have the ability to predict failure. Can you go into a little bit of detail on that, even on the technical side? Sure, so here's what I'll geek out a little bit because I am the techie guy. Let's go for it. So ECC is known, you have to have ECC to correct for data errors. It corrects up to a certain number of bits of error. We can do 32, 64, whatever you want to call it. But going beyond that, NAND has an inherent wear out characteristic, so it has program array cycles. There's device physics involved where you're actually wearing down the oxide layers, the silicon layers inside the chips. So they can last today, an SLC device lasts 100,000 program array cycles per bit. And MLC lasts 10,000 program array cycles per bit. What you do is you have the ability to know exactly when that 10,000th cycle is going to occur. So you can tell the end user, hey look, my drive's coming close to its end of life. The real potential there is what you can do with that. Can you extend the number of cycles by using proper controller technology? Can you move the data around within the SSD to extend the life of the product? And we have technology we've built into our controllers. One of them is a proprietary algorithm called CellCare. Cell because they're individual NAND cells, where we can take an MLC drive at any given capacity and make it last five years. So you can beat on it all you want. You can rewrite the capacity of the drive 10 times. So if you have a 200 gig drive, two terabytes a day can go in and out of that drive and it'll last for five years. And that's something that we can do with our unique algorithms in the back end of these devices to help promote customer use. So they come up and say, well, I hear it only has 3,000 program array cycles. Well, the raw flash, the way it shipped, has 3,000 program array cycles. Our brains, our controller, can take that 3,000 program array cycles and make it disappear. You can go 50, 60,000 times that to give you those five-year lifetimes. And so yes, they have predictable failure rates, but we make the predictable failure rates so far out, nobody's going to care about it. Well, the key there is you do it with software so you don't have to throw a bunch of flash at it. Yeah, so there's, exactly. Too expensive to do it that way. Yeah, so there's all kinds of interesting things. I mean, we've been lucky enough to get about 83 patents towards technology inside these devices. Three is good enough. For example, the 146 gig drive had overprovisioned down from a couple 100 gigabytes. The 200 gig drive we have today is 256 gigabytes. We do hide a little bit. There's about 50 gig there that we call a, it's been compressing. It's been termed overprovisioning in the industry. Everybody's really doing it now. Any device that comes out that's an enterprise-class product has some form of a provisioned space, and that also helps add to the speed capabilities of the devices. Oh, sorry, I thought you had a follow-up, but... Sorry, I was drinking some water. And I wanted to talk to Scott about the whole storage hierarchy. So there's a scenario out there that says, all right, we're going to have flash and then what I call it before, the bit bucket is deep, it's cheap and deep. Right. Do you guys subscribe to that, or what are your thoughts on that? We actually, the tiering model is very sound. We've got products that fit into every aspect of the tiering model, and we figure out how to work with our customers about what plan they have for the system. So if you're doing a system data that's 15K is your high performance tier, 10K, and then, you know, saddest spinners for your whole storage, we can either replace the 15K tier, or we can augment the 15K tier. I've actually got a breakout session tomorrow where I'll be actually putting up an ROI model and showing the actual cost difference. By putting the more expensive SSD in the box, the overall cost of the IT user is still less, and the value add and the performance. Yeah, I believe that. And so, but your answer was very politically correct, as you're saying, basically. Yeah. Whatever the OEM wants to do, it'll be support, right? With multiple tiers. But so, or are you saying, no, we see that it's going to be multiple layers, that there's value. It's going to be multiple layers. And there is significant value, because I mean, I can give you an 800 gigabyte SSD today, and you can put a whole bunch of 800 gigabytes in there to give you your bulk storage. But the controllers, the back end of the overall system can't manage the performance that you're going to get from that many drives. One of the early problems we ran into with the first launch of the products was the controllers that were running the RAID groups or the expanders, they couldn't handle the bandwidth we were pushing out to, because they never expected to see that level of IO come off of a J-Bot or any kind of architecture. So, personally, I think you're probably right, because you can't just replace all those tiers overnight. But the follow-up I had to that is, there's granularity now in the IO stack where you're seeing you guys generally connect into an array. You're seeing people do PCIe and you're seeing stuff in between, flash on controller and you're seeing all kinds of software. So, what do you see happening there and is that an opportunity for STC? Absolutely, that's one reason why even today we have so many options available to our customers, is interface choice, how much you want, where you want to put it, how fast you want it to be, flash technology, do you want it to be moving forward? Because we now can support SLC and MLC in enterprise-class environments. So, by doing that, yeah, you can put it where you would like to put it. We've got solutions that work as from the hot, pluggable front end to the mount in the back. We've even got little small embedded devices that go in the back end of a 1U chassis. So, you take out the smallest hard drive you can get today, which is a 40 gig boot drive, put in a simple small SSD, takes up half the space. You can now cram in more controllers and cram in more DRAM when you need to on that smaller chassis, or go for the big box and put one of these guys in and get your bulk storage and your performance. So, the former was still emulating a spinning device? Yeah, they're all emulating. And then you have the move to things like PCIe architectures where you can get the PCIe interface, but there's so caveats there. We are in the process of getting something there. It was announced by our CEO during our last earnings call that a PCIe project is in the works. But there's trouble there, too. So, there's already players in the market. They do things very uniquely. Fusion IO IPO is coming this week, and then plus a violin just raised a bunch more dough. Exactly. So, they have their unique violin box. Fusion IO has a software-centric device, so they have a lot of middleware they put on their systems for it. So, how do you build a solution that makes the most sense for all your customers? And then what interface do you use? Because you can have that PCIe, now that it's PCIe, it's not really talking protocol like a drive would. Do you put a drive protocol in front of it? Do you put a man flash interface protocol? Well, they say it's got the best IO as no IO. So, maybe you don't. But no, to the point, I wanted to have a discussion about the increasing granularity of that side of the IO stack. And so, you're actually seeing the hierarchy become more complicated, not simpler. So, I don't necessarily subscribe to that flash bit bucket because I see it just being very granular for a lot of different use cases. Oh, and it'll fit in a lot of different places. Again, because of its architecture, the way it's built and how you can access it and how small it is compared to what you're used to for traditional rotating media or the cost of DRAM versus the cost of flash, it has a lot more applications and it will see a lot more growth in the more off segments that you call it, like the controller level architectures, the PCIe architectures. So, it's just a matter of how the OEMs choose to use it and utilize it. You can produce a product and ask them to use it your way or you can talk to them and make sure the product does exactly what they need it to. And our goal has always been as the OEM company that we are is to make sure we provide the product, the customer needs, not provide a product and ask them to try to fit it into their square playground hole. So, new form factors, a PCIe statement of direction, what else is on the to-do list or is that going to keep you busy for the next? Well, that'll keep us busy for a while, but what comes next is, okay, geometries are getting smaller, reliability is getting worse. What happens when you hit the theoretical, beyond the theoretical of Moore's Law, what's the next technology? So, you've got all kinds of new architectures that are coming out. Didn't Michael talk about something like that? Yeah, he was at. Some gaseous storage technology, I don't know. Exactly. So, you've got phase change memory, you've got 3D NANDs, you've got all these other things. We've got a whole set of engineers that are already dealing with and working with all those types of devices today, so. Got your best people on it. We do. We've got, I mean, my former experience as a semiconductor guy, I used to design NAND chips, so. Where was that?