 2011 we are on the ground in Las Vegas. I'm John Furrier the founder of silk and angle comm I'm here with my co-hosts. I'm Dave Vellante of wikibon.org And this is our continuous coverage of VMworld live 2011 and we're here with David Flynn the CEO of fusion IO and His new employee or soon to be new employer rich bowburg CEO of IO turbine a recent you guys recently announced the acquisition of IO turbine so welcome gentlemen. Oh, thank you Thank you very much. David. Thanks for coming back to the queue. It's good to be here It was fun last time. Yeah, so whirlwind for you. So first of all, I want to ask how's it feel very humble last year in the Cube telling you story predicting the future SSD is hot valuations are off the charts from the VCs They're investing in everything that moves you guys went public. So tell us. What was it like last year for you? last year was a whirlwind it really was a An amazing journey to get this far and it's really just the beginning I mean, that's that's what it takes to get to the starting line You know is to get to where you're a public company And it was just amazing to do that with such a great team along the way Yeah, so the we've as you know tracking this business pretty closely There's a lot of confusion About flash. Oh, yeah It's the Wild West I like to say because it's such a new technology that people haven't really formulated in their mind Kind of a way to categorize or understand the different ways to use it It's like everybody uses disk drives, but that doesn't make them EMC and net app, right? So everybody uses flash that doesn't make them fusion. I know there are very big differences in business models in architectures, you know in In how you know how you use the flash from a technical perspective how you interact with customers all those things Yeah, and you guys are really a systems company, you know first I don't really people are starting to understand that they got you know our system is a distributed system It has componentry that's scattered throughout all of your servers It's like taking a storage array dicing it into a thousand pieces and putting it in your servers But you can't go back to the days of direct attached storage where you're managing each server as an island unto itself Which is why the real challenge isn't just creating the hardware that's capable of housing large quantities of data With very rapid access the challenge is how do you now run a distributed system where the data gets Diffused to the right server at the right right place the right time and that's really where the acquisition of ioturbine comes in And we're so pleased to have Rich Boeberg and Vikram join us because they had a similar vision on On how to manage in a virtualized environment the supply chain of data well and rich I we've seen the you know as a former spinning rust person right We saw the hierarchy in the 80s and the 90s start to get pretty granular We're seeing similar things with flash now right we've got server-class memory You know all the way down to devices that emulate this disk so the ioturbine acquisition was very interesting because I think it's recognition that The you know the fusion IO classic fusion approach isn't going to be all things to all people Right and and so ioturbine comes in rich I wonder if you could talk a little bit about ioturbine for those who don't know the company You know Take us back to the beginning what your vision was and you've obviously seen that through to an exit But give us a quick snapshot sure so ioturbine was recognizing the need for Flash in the virtualized environment It's a systems business in the non-virtualized environment a virtualized server is an order of magnitude more complex It's instantly Systems business on steroids our approach was to go after the the IO performance issue not the the CPU Hypervising or the memory hypervising but the IO performance Hypervising and we needed software to bring flash into that server and to address that that that basic bottleneck in the In the hypervisor world in the virtualized world so so we've got this hierarchy developing server class Flash yep, let's call it with VSL You're talking about eliminating the whole we view that as kind of a hypervisor for data It houses data from multiple origins virtualizes that the same way the VMware hypervisor hypervises the computational resources And memory resources and what's so cool if I can you know Talk up what they do sometimes it's harder to talk about what you do It's just really cool what they do is they actually have built the para virtualization Infrastructure to capture the IOs before they even go through the guest operating system and to tunnel them down into where they can get access Through the VSL into the flash and surf so what this means is you can actually serve Ios in a way that bypasses not just most of the the the ESX hypervisor IO stack But also most of the guests actually captures them above even the file system So it creates that that para virtualized tunnel for the IO down in so you guys love and running overhead Noticing that we are all about high potency flash How do you get flash to render the most benefit because it costs on a per-unit capacity so what you want to do is Accentuate what it can do and and this the results have been phenomenal with our you know beta sites What we've seen is between three and six times the workload per server now That can be used to put three to six times as many virtual machines on that server Or in some ways even better is to get more workload on each individual machine So now instead of sitting there idle most of the time your Microsoft SQL server your exchange can get get more throughput Right so no matter how you boil it down when you can do six times as much workload on the server It's cutting to one-sixth the effective cost of the entire equation So let's break down that hierarchy and we've put this out on wiki bond I don't know if you agree or disagree with it But at the at the at the lowest level I talked about disk emulation SSD like a stack drive and an emc device That's kind of for legacy applications That's right Right and and then then we're seeing this sort of flash cash You see that in hardware and software and different different combinations and and that's really we're more where IO turbine fits Yeah, if you think about it We're trying to treat flash more like memory than like storage than like a disk drive And you wouldn't put your DRAM on the other side of a storage network You would want your DRAM right next to your applications. So that's where the flash can really be used It's two orders of magnitude faster than spinning disk And so just the latency going across a storage network would slow it down So you really want it right next to the application and what we do is cash the IO Accesses really the dynamic aspect of the storage next to the applications. We leave the the Primary storage the big storage systems as the repository for the static image of the data But if you really want to solve the performance issue you want to move that closer Yeah, and cost wise you're not going to compete with you know sad a bit buckets. No, it's not for capacity This is not for data at rest. This is for data. That's in motion But but then you've got this all-flash block based storage that's targeting like you know the classic You know Symmetrics and IBM DS Well, you know people have always used RAM Quite effectively to speed up those storage arrays to be so flash really doesn't stand to have as great of an impact on The storage array itself because people already use battery-backed RAM for the for the cash What really the you know the bulk majority of the RAM market though has been in the server All right memory used in the server is is always more valuable than on the far end of an hour It has to do with just the the shipping costs You know you don't want to you don't want to have to send a ship to China every time you need a new piece of material for your for for your Assembly line, okay, and then at the highest level then you got the fusion IO And of course you take us but in bringing via salad now you're eliminating disk protocol overhead That's right We don't have to go through microcontrollers that emulate disk IO and the higher you go up that hierarchy the more expensive It is but so you're talking about getting value out of those applications, right? So that is that a reasonable way to look at the landscape out there Yeah, you're playing in now a big swath of that with and it also ties into our Our ability to support enterprise customers with a direct sales and support team It's kind of ironic that My customer support team spends more of their time fixing the other stuff that breaks when you push servers to do This kind of workload. I mean it's it's amazing that you can get Six times the amount of usable work from a server it speaks to the fact that they're sitting there starved for data Most of the time, but they're not used to running at 100% utilization David talk about your Competition and then your your customer base. I know your public companies You really can't go into too much detail on the customer base But since you've been so successful you retracting competition across the board obviously direct competition and indirect competition Talk about that and then talk about your customer base and how that's changing and evolving as you're growing Let's take both those questions first competition We compete on four different spheres because we're a systems company, but this is a new market So we're vertically integrated because the pieces up and down the stack didn't exist before So we started with building a hardware platform attaching NAND flash like a memory Which to do that required building software to teach the OS how to use it. That's VSL Those form the kind of the foundation like the hypervisor for the data piece but at the same time we were building the The sales and support and the ability to sell solutions directly to end users in creating this new value chain establishing new partners with the OEM server vendors with the OS Application vendors where all of this has to be embedded inside of the OS And so we view these as kind of four concentric circles the hardware itself where we excel But it's actually kind of the least interesting It's the software stack within the OS to virtualize flash as a memory and then it's the systems level software that runs the distributed system That's where IO turbine fits in direct cash fits in IO sphere that manages the distribution and flow of data but probably more important than all of that is the Sales and go-to-market and business model of not being a component vendor and not being subject to somebody else's route to market because the The the realization was that existing systems vendors are kind of compromised The value proposition that they're used to selling has a different formulation and this is in conflict to it And so we said this is an opportunity for us to take a value proposition Right to the folks who benefit from it most and not be conflicted by the fact that you know storage guys want to still Spend or want people to spend a lot for premium storage Server vendors want to sell a lot of servers So it's you know, it's kind of how do you how do you do that? You have to get in there and make the case yourself with the end users now two customers obviously Facebook and Apple or 10% customers you know Big cloud players. Do you see those guys as as outliers or is that sort of the future of the data center? So our customer base is a full spectrum Very few companies rise to the consumption potential of Facebook and Apple that's why they stand out but in reality it's a spectrum kind of a pyramid You know, there are deals in this case that are in the in the many tens of millions of dollars There are deals in the millions of dollars there are deals in the hundreds of thousands of dollars, you know And at the bottom end of the spectrum, you know, we have customers who? Would have had to been facing deploying an expensive storage array and drop in a card for a mere ten thousand twenty thousand dollars and No longer need to go to an expensive high-performance storage array to support their infrastructure So the spectrum is huge those two are simply the the tips of the iceberg that poke up, you know from from the foundation but we have over 1500 customers to date and that's squarely in enterprise business largely around database But also unstructured text and basically anything that needs big data with virtualization now supported We can even address those applications that aren't inherently data intensive but become such when you bundle them together and This allows you to remove that bottleneck so you can consider Continue to consolidate that makes it not just a performance play But also a cost effectiveness and consolidation play. Yes, you've got a big tan But do you see those those outliers those huge companies as Indicators of the future or do you see? Oh, you bet. I Mean those guys are the ones that are looking at how do they run their infrastructure in the most efficient way possible? They're the ones that stand the most to gain from understanding these things and it's it's it's fun working with them because Because they are out there and you know with the with the economy the way it is recently It has let more companies to go and look is there a better way to do it? Is there a more cost of way effective way to do it in good times the adage? You know, nobody gets fired for buying IBM that holds true in bad times when budgets are tight You might have to lay somebody else if you want to continue spending about the same That's about innovation. You mentioned innovation about the product in the business model It's very Apple-esque and when Steve Jobs announces resignation Steve Wozniak went on Pierce Morgan and Talked a little bit about it, but he couldn't stop talking about fusion IO and you so how do you feel about that? I mean Steve Jobs Apple legend and the was praising you guys big time so Talk a comment on that. Sure. So first Steve Wozniak is an engineer's engineer He that's the highest calling in life and he is an amazing individual when it came to our engagement with Apple He stayed out of it. He didn't want to be involved, you know Because of optics and you know how it might look and so forth. So he really didn't have anything directly to do with it But you know he came and fell in love with fusion IO the two questions He asked when we first sat down a you know series of coincidences led us to having lunch and And he said okay two things I've loved solid state moving electrons makes a lot more sense than moving atoms And so I've always paid premiums to buy you know the newest and greatest solid state stuff so why is solid state going mainstream now and What is it that makes you different and the answer was Solid state is going mainstream now because NAND flash has the ability to have a hundred times the capacity density as Ram because it doesn't give off power and it's natively higher density So the only reason we're here is because NAND flash makes a better memory Making a faster disk you could do that with RAM all the time right? It's the fact that it's a higher density memory a good enough memory that actually Has some advantages in being non-volatile that was the answer number one is NAND flash Displaced RAM has much higher density today because of consumer electronics Apple driving the volume of the flash right so then the The second thing was why are you different and we we told him that we looked at it from first principles and said if storage as we knew It today if the data supply Methodologies that are used today didn't exist How would you employ flash and we came to the conclusion that it's two primary things you would attach it? Like a memory and not go through the emulation of disk drives using microcontrollers and number two That would lead to the need to run a distributed supply chain of data This is like Walmart Walmart Runs a supply chain that's very efficient because they have local distribution hubs their distribution Efficiency is what makes them as a company This is all about distribution of data within the data center and you're not going to do it from shipping everything from the same Central location every single time and so that that inversion principle is what really led us to think This is going to be a sea change and the amount of technologies are going to be developed to manage a distributed supply chain We're going to be immense Awesome David Flynn rich Bober great story great story. Thanks for coming on the cube again David Flynn The super techie turned CEO love that we love that that path and rich Bober old friend You know glad to see everything worked out great for you guys Congratulations on the IPO way to go in the first quarter. You know now just do it a hundred times more in a row and Thank you