 6 o'clock and Gary is no stranger to us programming you were on at VMworld, you're on Astrata, you're here, we keep running into each other, you're awesome, we love you, you've been on the Cube multiple times, you're a writer, you're an executive at Fusion I.O. and you've done a podcast, you're a video star, you're an executive at Fusion I.O., you're in the storage business, that was this last week, what are you doing today? So today I'm here with Fusion I.O., we're sponsoring SNW, we have a big booth, this is one of our products, we'll see, I don't know if we can fit that into the machinery here in the video station. I'll do the whole Vana thing here later. Yeah, so the demo just to share with folks who might not have a chance to be here in person, the demonstration we're doing today regarding our Fusion I.O. drives is that we have this gigantic video wall and the video wall is supported by 20 servers, each server is serving 225 video streams, so that's a total of 4,500 video streams coming off a single Fusion I.O. drive product. Now that product is a little bit different than this one. Can we get that for this QOS for our bandwidth? Yeah. Come on, you guys are the ones seeing the connectivity. Well no, this is just local, this is just video streaming locally. So what am I showing here, we're going to go back to your thing. Yeah, sure, this is our I.O. drive octal product. Sorry, the I.O. drive octal is in the demo booth. This is our I.O. drive duo. So this product is a PCIe based flash memory card that can go inside a data center server and basically be used to accelerate databases and applications. It's flash memory based, which is obviously much faster than traditional rotating spinning disks. And the real fun part about what's happening right now is we're seeing this trend of taking the process critical data or the active data and putting it as close as possible to the CPU. We went through this wave, really part of the storage networking industry association, about centralizing storage external to the servers. First we had it internal to the servers, then we externalized it. Now we're finding that in order to get the performance that's needed, we have to put it back to the CPU. It also has to be in the same kind of silicon that we have with the CPU to match up the speeds and feeds. And you know, today's servers, they're dealing with more data. They're dealing with more connectivity. People are coming in through every manner possible, not only just computers but laptops and mobile phones. And nobody has any patience. We need everything right away. Yeah, yeah. And so these I.O. drives that we deliver to customers, they can place them in their data center servers, speed up their Oracle database, their Microsoft SQL server database, or use it for virtualization. VDI, as you guys know, is a huge hot topic these days. And we're finding a lot of folks who are able to deliver more virtual machines per server using this kind of product. So that's, you know, a quick overview of what's- But it's interesting the trend you're talking about, for a decade and a half we pushed function out to the other side of the channel and now it's coming back. Exactly. But the difference is this time it's persistent. This time it's persistent. So what does that mean? Well, let's go back. I mean, we'll do a little quick history lesson because I think it's kind of fun and, you know, all this started at the Storage Networking Industry Association, oh, I don't know, 12, 13, 14 years ago. 1999. So what happened is we had servers with disk drives inside them, and that was the model. And the way to get performance out of that infrastructure was to use more disk drives, string them all together. But if you need all those disk drives in a single server, you can only fit so many in that server. So, you know, part of the impetus for externalizing the storage was performance because once it was external, we could put dozens or hundreds of drives together, have that performance, and share that across multiple- Tons I've all been with. Exactly. And so- And certainly there were other management benefits to that as well. Sharing. Sharing, right. But the idea was now we have that performance. Well, today, again, the world has changed a lot. The internet grew just a tiny bit over the last 12, 13 years. And so, more data, more connectivity, need for speed. And so now what we find is if the server is dealing with a request and it has to turn around and go get that from the sand that's back at the other end of the room, the latency is just too much of an impediment. So by now taking the process-critical data and putting it right into the server, using an IO drive, we're finding that we're able to get the performance improvement that people need to meet these kind of workloads. And what we're calling this is this is really- we wrapped this up in sort of data decentralization where you can- you know, we wanted some of the data that was centralized, but now we want to decentralize some just like we've decentralized CPUs. So another interesting analogy to think about here is with the mainframe era, all of the CPUs were in the mainframe. And we had these dumb terminals, essentially, at the other end of the wire. Well, when we went to client server, we kept some CPUs at the server, but we put some CPUs in the client. Desktop computing. I mean, companies like Microsoft coming up and all the benefits of the applications and the rich experience that we got. Well, now the same things happening with performance oriented storage. We don't want to keep it all locked up inside the equivalent of a storage mainframe and have the servers have to do these acrobatics to go get it and bring it forth. We want to put that process-critical data right into the server. So we sort of refer to that framework as data decentralization. You talked about the latency. Now that's two-fold, right? Because not only is the disk spinning, but there's all kinds of overheads associated with it, right? So in other words, why can't I just shove in a bunch of SSDs into an array, for example? Well, it's technically possible to do so. But one of the things that happens in that scenario is you're dealing with a lot of legacy storage protocols. So most frequent ones that we talk about are things like SATA or SAS. SCSI. SCSI. And so those protocols were designed for rotating media and did a great job. SCSI actually was designed for tape, I think. Before that I have to check my history books. But in the world of flash media, that doesn't always do justice to what flash memory is capable of. And so what we advocate is let's put flash memory in the best position to succeed. Let's allow it to put its best foot forward. And doing so means let's not let it hide in the back of the data center behind all these interconnects and all of these protocols. Let's bring it face forward, process critical data close to the CPU because the goal here isn't necessarily just to make the storage faster. The goal is to keep the CPU humming. That is the key cog in the wheel and the utilization that we have to keep going in order to serve the data that we want. More data, more users, more quickly. Multi-core is a big factor there too, right? I mean, a lot of IOs coming out of Intel these days. Exactly. And so as the server platform scale in terms of the number of sockets that they can hold, CPU sockets, and then the number of cores, we get this great scalability at the CPU layer. And we don't necessarily want that all to go through one cable out to a storage array. How about having that available to the entire PCIe bus on that server? So if that server might have four, eight, ten slots for PCIe, you can fill those with FusionIO drives and get a really great parallelization between the CPU bandwidth and the flash memory. The RAM is too expensive and it's volatile. So now we're seeing a hierarchy emerge, John, with a whole new set of applications, right? Yeah. Which is just your sweet spot, right? Yeah. So obviously you're here. Gary Arnstein, who's the vice-president of the product market for FusionIO, innovator in new kinds of storage, data decentralization, or storage decentralization, whatever you want to call it. You guys are going public, so you really, you know, a lot of hype going on around FusionIO, a lot of interesting data coming out from the competitors knowing that you guys are, you know, our hot company. And I wrote a post about this and I want to just ask you this. And that is that there are new companies out there that are in these new franchises, the Zingas, the Facebooks, the Twitter, these new applications that are seeing massive scale. Massive scale. That are relying on things like very intensive database-driven type applications, custom open-source coding, Facebook and Twitter. You know, it's documented in their history about how they're home-growing, a lot of stuff, and operating at scale. And so you guys, you know, sell into these kinds of environments. And so, you know, there was a post going around, around the ideal market for you guys. And one of the things I said was, it's not about selling into Facebook as a per-unit thing or selling into Twitter as a per-unit thing. It's about how many environments are going to look like Facebook. So the question to you is, what's going on in the marketplace? Are more companies going to look like Facebook, or more enterprises going to look like Facebook? How does a Facebook or a Twitter or a Zinga, people who are built on cloud, built on essentially high-scale data centers, using cutting-edge techniques like SSD, how are they a proxy for what the future enterprise will look like? Yeah, I think that's a great question. And I think we have to look at it. Certainly companies that are operating at that scale with those kind of applications are a leading indicator of what might come in certain other markets. But at the same time, the applications of a company like a Facebook or a Zinga or a Twitter or something are very different from some of the classic enterprise applications. So I'm not sure it's a one-to-one type of relationship of if one of these large, web-scale companies is building their architecture like this today, we're going to see that in the enterprise in two years. I don't think we can draw necessarily such a... I've done a one-to-one. A philosophical mindset. Sometimes a platform. They're using open source. There's a consumerization of IT trend that's being talked about here that's pretty heavily. Mobility and cloud and... I think what is interesting is that the change in the amount of data, the amount of connectivity, and the need for speed is universal across these different markets. So whether you're in the web-scale market or you're in the enterprise market, everybody's dealing with those same three issues. And so we might solve them a little differently. We might solve a solution for an Oracle database a little bit differently than we would solve a solution to serve millions of photos to millions of users. But the underlying issues there are common, which is that we can no longer rely in today's world more data, more users more quickly of going back to this rotating mechanical media in order to serve the data. I don't know if that answers the question. We asked some of the people at the Stratoconference, the top markets, and we thought, I thought web would be, large-scale web apps would be, or platforms would be number one. Yahoo, Facebook, Google. But apparently, government 2.0, military, healthcare, finance, these are verticals that are exploding utilities. It's not just Facebook and Zynga and Yahoo, it's these other verticals. I'll tell you one of the most interesting things I've seen around is companies that are dealing with literally three workflows in one, which is they have to capture a whole bunch of data, often in real-time, so if you think about things like capturing weather data or capturing some kind of sensor-related data, so there's that incoming stream, then of course everybody wants to visualize or somehow make sense of that data in real-time because we all can't wait for half a second. And then the third thing going on at the same time is they likely can't keep all of that stuff that they need, how they're going to archive some of that data. So these three kind of workflows that capture the real-time analysis and then the simultaneous deciding of, you know, how am I going to archive this, that's a really unique problem that you just can't solve with traditional technologies and you need this kind of approach of take that process critical data, bring it right into the CPU, active data close to the processing, keep the processing humming, and deal with these workflows simultaneously. The other thing is that I think that, you know, the web stuff is very interesting and a great indicator of what's coming, but the enterprise market is really adopting this kind of technology en masse and we see solutions. What are the drivers? What are the drivers for that? The drivers there are databases and, you know, who doesn't want their database to go faster, right? So keep in mind, you know, we talk a lot about applications but the applications are supported by databases, so whether that's Oracle databases or Microsoft SQL Server or MySQL, we have the MySQL conference coming up in Santa Clara, I think it's next week. That's a tremendous area of interest. The HQ opportunity. The other, yeah, we'll see you there. There's no rest for the weird. Sorry, Deb. The other area is virtualization and particularly we're seeing a lot of interest in virtual desktop infrastructure. One of the neat things about virtual desktop infrastructure of course is, right, you can allow everybody to sign in and get that virtual machine on the fly. Yeah, provisioning apps. But if they all come in at the same time, Big BootStorm, yeah. Big BootStorm. So the very simple architecture for giving a boost to your virtual desktop infrastructure is to take that golden master and to take those clones of that operating system and put that on flash memory products such as a Fusionisle and boom, you know, all those problems go away and you can support more virtual machines per server than you could before. What do you think about VDI, virtual desktop, you know, desktop virtualization? I think it's here. You know, it's one of... It is... Isn't it not... It's been talked about, but now we're kind of on the doorstep of mix happening. It's the real thin client, right? Isn't it? It's the final instantiation of what we were all hoping to see with just... I don't know. I'm not sold yet. I mean, I'm happy to be... Why? What makes you not sold? Because I don't want to give up my desktop experience, you know, and I think most users don't. So if I'm a claims operator... Is an iPhone a VDI? Is a tablet VDI? Well, not today, right? Yeah, but if you're working in a large company, right, where you're going in and you're logging in and, you know, it's so much simpler in that environment to provision these virtual desktop... It's simpler for IT. I totally agree with that. It's simpler for IT. But from a user perspective, I don't want to give up that user experience and so that's why... What would you know? What would you know? I mean, I think we're getting to a point... Would you feel like it's transparent today to the user? You know, we're not using virtual... I think if you're a claims, you know, a jester, if you're working on the claims desk all day or you're doing, you know, a call center, yes, no, it fits there. But I think for the... What you were talking about, Gary, about the thin client vision, the Scott McNeely vision of thin client, I don't think we're there yet. I think... I don't think we're ever going to be there. You never have to take away hard disks from somebody. They're going to have hard disks to play with. And I think it... Because the economics aren't there for a virtual desktop. Yeah, but the economics for companies that are deploying this clearly are there, right? Yeah, oh yeah, oh yeah. For those use cases where it works, it works great. I agree with you. So that's where we're seeing a lot of users, right? For companies that want to deploy lots of virtual desktops to lots of users, they're able to support, you know, more per server. I think I agree. I think I'm going to go Gary on this one. Because I think it's here. I'll tell you why. I think that, you know, back in the PC era, back in, I'd say, early 90s, the notion, you know, when the 46th kind of hit, that team was right on the next gen. You saw a mass distribution of PCs and every distributor or reseller at the time was basically doing, basically burn disks because the resellers didn't want to load windows on every machine. So, you know, from a bottleneck standpoint, the dealers would, I remember the days, you know, they would be like, okay, Fidelity Investments or XYZ wants, you know, 500 PCs this month. Guys plugging in a freaking CD-ROM pumping windows on it. So say, hey, master disk, there are all these hard drives and this install the hard drive. So I think what he's referring to is, on an enterprise, everyone's desktop will have some VDI-like app. Right. You might have a hybrid, too, right? You might have a local desktop application or environment for the people who need that. But the ability to pull up a virtual desktop environment. So let's talk about what this means for Microsoft because Microsoft freaking hates desktop virtualization. Yeah, they should. It's really disruptive to Microsoft. I mean, they're doing a lot of promotion around hybrid V. Of course they are, right? Yeah, absolutely. But so when desktop virtualization first started, you know, to come out, Microsoft's now poo-pooing it. Right now, of course, Microsoft's embracing it because they don't want to... Because no one's upgrading Windows 7. Well, so... Maybe because they like it now. But my point is this, we're going to test the lock-in that Microsoft has. And I actually think Microsoft has plenty of knobs that they can turn to keep the VDI camp economically less feasible. They'll just keep ratcheting and now changing the licensing, you know, keeping those customers locked in and keep them extracted. It'll be interesting to watch. So lock-in is now passing. Right? So would you agree lock-in is passing? Well, let's say... Or has it ever been involved? Well, it's horrible in Microsoft. Well, it depends if you talk about from a vendor perspective or from an end-user perspective. That's always been the end-game. Yeah, vendors love lock-in, right? I mean, come on. Listen, like lock-in. But tell me, VMware doesn't have lock-in? I mean, they have a degree of lock-in, right? Yeah, I mean, it's all about choices, right? So does Google have lock-in? Do we have more choice than we did 10 years ago? So here, this is a digression, but it's a fun one. Was the world better or worse when the only choice was an IBM laptop, Windows, BlackBerry, and Microsoft Exchange? I sometimes long for those years. Because of the simplicity, right? It's so complicated. I only had a cell phone in my car, which is great. I played golf and did some things. It was not a phone call. So I want to switch it up. What are some of the interesting things you've seen here at SNW this week? So I just got here about an hour ago and came right into the queue. The most important thing that's happened here at SNW is that the queue is here. So we are here live. We're validating SNW. So SNW has finally stepped into the modern age of media by having us here. I've been reading some of the stuff that I've seen. There's a lot of cloud stuff, right? Yeah, there's a whole cloud exhibit here where they did an area where, you know, folks had a, there were a collection of smaller setups. I'll give you my two cents on this. Dave, just because it's a cloud infrastructure. I don't even need debriefed yet, but my angle is this. The era of innovations is obviously on SNW. They're using Silicon Valley as the kind of place to kind of domicile around innovation. Is this the first time that SNW has been held in Silicon Valley? I was told that in the old days it was held here, but certainly for the first time in the last five or six years. So it's an old show. You guys know it's my first time, and I know about the big whales that were here. But in essence, you know, storage has changed. It's changing, right? You guys, you know that. You're a fusion IO. So innovation is the theme. I think that's just more of a kind of a bumper sticker to get people pumped up. But the theme I've heard from people is twofold. Cloud and SSD. This is about flash. Flash memory. I've heard a lot of people flash this, flash that. And David Floyer from Wikibon and I were talking at lunch. Performance is a hot one. So to me, this is coming out of the underlying thread here is performance, performance, performance. Performance is important. But you know what I like to share with folks is performance is often the thing that drives people to these kind of solutions. But it also has, there has to be a kicker for cost savings, right? Performance without some cost savings is just, you know, just sort of off for having fun. And so what we find with our customers is that, yeah, performance will bring them in the door, will attract people to a particular solution. But if they don't immediately see some cost savings in terms of consolidation of infrastructure or the need to postpone the purchase of additional infrastructure, some hard dollar savings, then it's no longer interesting. So it's really a one-two punch. It's can we reach some new threshold that's going to allow us to do more than we were able to do previously? But then two, can we pay for that with the savings that we're going to get by using less equipment or not having to buy new equipment? Yeah, yeah. So people who have a database performance problem throw spindles at it, they throw compute at it and they don't solve the problem is what you're saying, right? Yeah, I mean we have a number of case studies where people talk about, you know, I was trying to add more spindles to this configuration. That was, and that's been before, you know, flash media came onto the scene. That was the recommended method for achieving more performance. So people try that and then realize, you know what, this really isn't getting me where I need to go. So I need to take a major step forward. It doesn't scale the way they need to scale. It doesn't scale the way they need to. It's very hard to configure disks for performance, very hard to configure disks for IO operations per second, which then translate into transactions per second for a database or an application. So they'll say, how do I move off of that? And yeah, in the meantime, all this infrastructure for housing all of these disk drives can be repurposed. And so that is a major cost savings from a dollar perspective. The thing that, it's a common thread thing along those lines is that two things are happening. Everyone is trying to think about what's next while it's trying to figure out what's happening today. Right. That's the comment we heard from the entrepreneur earlier. But to me, the common thread on that equation is economics and value add have to be tied together. So it's not just an economic game. Hey, yeah, save money. Sure. So whether it's cloud or whatever you're reporting to this new environment, this innovation that's happening, ultimately it's about change, right? So change is the key. But the change equation is economics and value add. And it's an and equation. Economics and value add. Yeah, you can't just have one. So that's kind of the common theme and all the best companies we hear from you guys and on is that the value add component has to be a major kicker. It can't be, oh, a nice to have. It's like big value add. Whether it's obviously database performance for you guys or something else. So that's kind of our filter. Right. Absolutely. So Dave stepped out. So we've got to wait woefully behind schedule here. OK. Well, thanks very much, guys, for having me on. It was a pleasure. Thanks, Gary. Always a pleasure seeing you. As always. And you'd be able to put this to use in the video machine? Yeah, I hope so. Yeah. Thank you. Can we get some of the screens? OK, we're going to bring Scotch.