 Trap below 400 that was okay. We're back. This is the cube. This is the flash flash cube This is Dave Vellante. I'm live here in Massachusetts and my co-host and colleague John Furrier is in our studio in Palo Alto Gary Orenstein is in the house. He's the senior vice president of products at Fusion I owe Long-time cube alum dating back to his giga-ohm days Gary good to talk to you again Thanks for having me and having us on the show a John appreciate it Yeah, always a pleasure. So what do you make of all this action and flash all the big guys getting in IBM's made You know a billion-dollar investment EMC made a lot of noise sort of aim the EMC gun at you guys. Well company F. I don't really know who that was but You know, you must be excited. It's getting it's getting interesting Gary. What's your take on all this? I think it's great It's in only just a couple of short years. We have gone from what some perceived as just a point technology You know off in the corner of the data center to something that is now pervasive across companies both large and small across the enterprise market in the hyper scale market and You know, it's hard to have a day go by without somebody making some big news about the impact of flash on the data center on application processing and you know, we don't see these times that often in you know, following the technology trends and I think we've seen now by the activity both the volume of the news as well as the products and customer solutions that are being delivered that this is a Tipping point of how folks are going to architect for the future Yeah, so you mentioned hyperscale John Furrier John you and I have been talking about hyperscale for a little while now and John I know you're seeing some major trends there. What maybe comment on that a little bit and then we'll kick it over to Gary Yeah, I mean Dave as we talked about in the Cube also We just said OpusDak last week talking about the developer angle around building clouds and you know Gary and I don't know each other when the whole cloud movement was starting and it was just a small bunch of guys This is before you know platform as a service took off and what you saw was the beginning of a movement where people wanted to build high performance high-scale infrastructure and that was mainly driven out of the demand for developers to essentially eliminate operations But also the beginning of the hyperscale movement and you know, we've had Amar Al-Wadala on from founder of Cloudera Who was at Yahoo before and Jeff Amar-Bhakar again another co-founder of Cloudera? Was at Facebook built their data groups These web scale companies really set the standard for what is now hyperscale in the enterprise And you're seeing that massive shift from open source to you know scale out commodity or industry standard hardware where you know you got on one end of the extreme you got Google building their own stuff from down to the bare bones metal and then you got Facebook which is simply assembling their own since you're using commodity parts and right in their own software So the big guys like Apple computer who kind of were pre hyperscale generation But then had to retool on the fly to support that massive growth So you see those kinds of companies setting the stage for the Hadoop movement for this new notion of using free software to really You know save money and spend the time on engineering new solutions and these new solutions are centered around flash So so you know, obviously, you know, we're bullish on hyperscale open source We think there's some compelling advantages on the cost side and also on competitive side with you know things like we saw Netflix Announcing massive earnings. I'll see the success of House of Cards And you know as we talked about at the Cassandra Summit with the Netflix guys You know they were running the only flash instance on Amazon web services Clearly a competitive advantage So it's it's Gary said moving from a point solution to a much more mainstream Competitive architecture and now reference architectures for those web companies are moving. So Gary my question to you is Obviously those web companies set the stage for this new migration where there's no more tax I mean used to be called the technology tax no more sun servers No more software now with open source that got the developer movement now that whole another tax level of conversations Moving into the enterprise where people saying hey, I don't I don't want to pay for Oracle I want to get off Oracle or I want to get off the VM where licenses I want to get off these these commercial software taxes To look more like Facebook look more like Apple and Google for instance So but now the enterprise can get there. So I want you to comment on the trend Do you agree scale out open source as a key driver? And what are some of the things you're seeing where the web companies are now showing the way for these for these Enterprises or these legacy search providers crossing the chasm so to speak and what are the dynamics? Yeah, well the fun part about the hyperscale market to me is that it starts with us as individuals We all have multiple mobile devices and we all demand access to data. That's anything anywhere anytime and When you know people ask me about hyperscale to try to put a frame of reference or picture mine I say well think about billions of people with billions of devices all Accessing applications that are going into the same data center because if you're using some shared application like a social network or a picture Sharing or something you're going to that same place that gigantic funnel is putting so much pressure on the data center To serve data more quickly now It's also putting immense pressure on the companies that build those data centers to do so really economically I mean how much do you pay for all the web services that you use? You know some you pay for some you don't So in the cases when you're not necessarily paying a tremendous amount of a subscription or premium their objective is to deliver as much data as quickly as possible but also at the lowest cost and Infrastructure becomes a core competence and as we've seen some of these large companies have said you know I'm going to shy away from the proprietary boxes of yester year maybe even last century and I don't care so much about badges and bezels in any of that stuff What I want is how can I achieve the maximum number of transactions at the lowest possible cost now? Transactions can also be different things transaction could be a bank transaction the way we think about the term classically a Transaction could be posting a status update a transaction could be posting a photo a transaction could be watching a video All those things are transactions that are happening at the back end application and so what's important now is this relentless pursuit of More transactions at a lower cost and we've seen that by using flash memory and Using open-source software companies are able to achieve things that were previously impossible Serving so many users at a price point previously unheard of You know holding so much metadata to index all this information that they couldn't hold before So those are a few of the big themes from my perspective. What are the examples? Can you share with us? I mean first of all? I mean obviously it has always been slow and they've always spent money to save time So they didn't have the expertise they outsourced everything now that migrating over to the hyper scale marketplace I've always on mobile Don Bacille your competitor was just talking about these new software models that are emerging on top of flash So can you want to talk about these new architectures and what are you seeing for solutions being deployed today in companies? migrating over to this hyper scale market and to What are some of the things that enterprises and IT guys need to do to move there sure You know so some of the most common deployments that we see I know very common one is just my SQL right my SQL's the world's most popular open-source database just about every company that is Supporting any kind of web application uses some aspect of my SQL and we have dozens of case studies of customers Large and small alike who have been able to shrink the footprint of their my SQL deployments you know supporting a voting application on 30 servers before you put in the fusion IO products going down to two servers After you put in the fusion IO products So that kind of consolidation in terms of the server savings for these companies is Enormous now for the folks who are using Commercial software they get can get the same benefits to right of where they previously might have had to have a certain size Deployment in terms of number of nodes. They're able to shrink that number of nodes and get more value out of each Individual server and each individual license for the applications that are at use now in terms of the deployment models. I Think now we're seeing the recognition that there is no one size fits all in terms of flash memory deployments and We focus on Really all deployment models for flash You can use flash inside your server as locally attached doors to provide the maximum amount of performance You can use flash inside your server and then software to transform that into a cache So you could Interoperate with an existing storage system and that's very popular in the enterprise market where they're looking to enhance For example virtualization solutions I don't want to change out their infrastructure so they can deploy flash as a cache They people do that a lot with our ioturbine software or folks want to deploy flash in a mechanism that they can share that across multiple servers in sort of a server flash appliance View and we do that with software that we call ion or ion data accelerator And the nice part about it is that across these deployment models fusion IO provides an open software-defined platform So that customers can choose whichever deployment models right for them We don't think there's a one-size-fits-all But the flexibility to use the same underlying platform in multiple deployment Architectures we find is very advantageous for both the hyperscale market as well as the enterprise market So Gary you you guys made an acquisition recently of a company ID 7 Some people took notice a lot of people were confused by it. Can you help us squint through what that acquisition was all about? Sure, so at the heart of many systems That serve storage out over a network. So let's go back to the very basics If you have a server that has storage resources in it and you want to share those out over the network You need to use What is called a SCSI target subsystem? SCSI being the the fundamental storage language that sits at the base of everything including fiber channel is you know Sits on top of SCSI and so this core piece of software with which essentially Allows the server to share its storage resources out over a network is part of the SCSI target subsystem that ID 7 built and they maintain an open source version as well as an ID 7 a version of that That's something that we've been cooperating on quite extensively For our ion data accelerator product and again, it's representative of this push To taking industry standard servers Infusing them with flash memory and software and transforming them into super servers We found that ID 7 and the work they've done with SCST was by far and away the most robust and best-performing implementation of that that particular need of sharing Storage services out over a network and you know after cooperating with them we decided to To take that relationship even further with the acquisition. Can you talk about the some of the things that you're doing? In terms of contributions to the Linux community in particular the kernel I think of you know the old days of paging and swapping and they sort of went by the wayside when the disparity between spinning disk and processor performance widened now that it's now that you In particular fusion IO and some others are trying to close that gap That Piece of the kernel becomes more interesting. Can you talk about that a little bit? Sure? I think it's also important to take a step back and recognize that we're out to improve the entire stack of Helping people process more data more quickly and more efficiently So that goes all the way from the underlying media to the the flash chips to the flash translation layer To the operating system to the applications There's nothing in there that you know is safe so to speak in terms of room for improvement and we have a number of open-source contributors at fusion IO who contribute to the Linux kernel as well as numerous other projects We're big supporters of that effort and there are as you identified cases where some of the mechanisms in In Linux are you know just were come from a day prior to flash being pervasive in the data center and so in this pursuit of you know getting rid of latency whatever the Layer in the equation. We've spent a lot of time and a lot of energy Contributing to Linux contributing to other open-source areas in order to to basically streamline the stack I think the important thing here is to remember we often talk a lot about This is being a sort of a storage centric effort and I I don't think that's always true I sometimes joke with folks that this whole initiative is not about making storage go faster It's really about making sure the application doesn't go any slower Yeah, it's subtle, but that twist and that mindset Allows us to really leave no stone unturned We don't have a legacy investment in Older storage architecture that we have to protect or propagate We don't have investments and legacy raid controllers and legacy sass and SATA storage devices Our objective is very clear in helping customers process the maximum amount of data with the least amount of infrastructure Both the hardware infrastructure and the software infrastructure as another example We're doing work with our software development kit to help reduce the amount of code that's needed to develop You know certain storage features and applications. So with my SQL as an example by the way the MySQL conference is actually going on as we speak down the road in Santa Clara We've introduced capabilities like atomic rights that essentially Handled the atomic transaction for my SQL. What's the benefit to the end user? You know, it can be as much as a You know 2x performance gain you reduce the number of rights by half You double the endurance and lifetime of the media. So we're helping increase the performance. We're helping Make the solution last longer We're helping streamline the the code So again, all of these things in this effort of squeezing the very last bit out of the hardware and the software infrastructure Yeah, now you talked about atomic rights We were talking by the way we agree with what you said at the top of this broadcast We were talking about the the limits that that spinning disk puts on on application design It's really about as you said, I like the way you said it not allowing applications not to go any slower But also there's such great potential to Bring new applications to market. We had Brian Bukowski on earlier from aero spike You guys are obviously the only one in market with the capability of atomic rights And then we asked David Fleuer about NVMe and you know, basically his his short responses It's stuck in committees. So you talked about some of these legacy, you know Raid controllers and vendors However, we would expect that you know, people are watching your moves and clearly others see the opportunity to develop capabilities not unlike what you have and try to leverage Their file systems expertise their OS expertise and maybe even put in an Infrastructure where you don't have to rewrite applications. We think of IBM in particular Potentially even someone like VM where what are your thoughts on that? What are you hearing in the marketplace again? Validation but also competition. What's your angle? Well, I think anytime you're looking at deploying new technology you have to straddle the Need for integration with existing systems and applications as well as the opportunity to break new ground you know, there's another joke I like to share which is Everybody likes change. They just don't want anything to be different And so we we have to do that with the the technology Advancements as well. So we need to develop solutions that will seamlessly fit with existing applications But we also want to make sure that we can carry the industry forward and sometimes that does involve change And the good news is customers can choose. Let me give you a specific example So when everybody introduced their first round of flash memory products Everyone made them look like a disk drive to the operating system that the flash appeared as a block storage device And that was a great idea because applications know how to use a block storage device and therefore don't need to change The difference is that most people not only made the product look like a block storage device They also architected like a block storage device and that's the one where I might have a difference of opinion from the way others have Have deployed because we believe that flash memory can go far beyond What disk drives were capable of and even are capable of atomic rights being just one example Some of the work that we've done to showcase other capabilities that provide much more memory like access to the To the media, you know, it's called flash memory. It's not called flash disk So we have to move forward while retaining the ability to integrate with existing systems and existing applications That's a balanced line that I think we always face in the technology industry John you were at the open stack summit last week and I wanted to wonder if you could Maybe talk about that a little bit and again bring the area into the discussion on the fusion IO's perspective on Open stack and generally and maybe even open source specifically Yeah, Dave, I mean, you know for the folks out there just trying tuning in We were at open stack summit all week last week for three days of live coverage And we essentially suck all the metadata out of that event talk to all the people there a lot of developers a lot of signal Not a lot of noise very relevant event and what really is going on open stack as we come into deal Greatly, but go to youtube.com slash Silicon angle. You'll see all the videos there I mean we did 37 interviews there But the main signal is people are building new infrastructure fast and it's based on scale out open sources We've been talking about but what's really critical is it's not just one piece of hardware or one port on a switch that's software-driven or whatever It's a operating system. So the element of flash is a building block So you're seeing architects systems architects all looking at cloud and looking at it from a perspective holistic perspective and saying hey I'm gonna build the next generation modern infrastructure and they're looking at a longer-term horizon and the arc of that Conversation is where my what do I need to do today to make my applications go faster? So performance is number one and two. I want to be future-proof. I want enough headroom so I can get more stuff down the road I want to read I don't want to rearchitect so the conversations are all about software software Architectures layering in using open source and using things like flash so to me I think what's relevant Gary is the flash conversation is about how do I build? How do I look at as a Lego block a building block? So I want to get your comments on how you guys look at that market obviously you guys are a key ingredient You're more than a hammer your your chainsaw for the old way and yet, you know big big foundation for the future So talk about the new architecture. Yeah, so what I see frequently Folks is in deploying new architectures They want to pick their favorite server or maybe one or two models of server They want to pick their favorite hypervisor or cloud platform open stack where a member of open stack big supporters there And then they want to fold services on top of that. That's kind of the model that I see I want to pick my server skews I want to pick my cloud platform or my hypervisor and then I just want to fold services in and I want those Server nodes to be very flexible in order to handle a database workload or web serving workload Or maybe a storage serving workload now in the conventional world once I had built out that Horizontal infrastructure of servers and I decide that I need to Spin up a database that can support hundreds of thousands of transactions per second in the old world I would have had to wheel some full rack of Sand equipment in there and sort of you know nudge it in between those skinny thin little servers to say I'm here Well, you can't do that with a line of code But if you've opulated some or all of those servers that are in your scale-out architecture with flash memory You can turn on a database to support hundreds of thousands of transactions with software or you can't slip You know something to be a web server that can handle a spike in traffic Or you can't slip something else to become all of a sudden a shared storage resource to support multiple applications So to me having the ability to have that infrastructure in place that can support Whatever is needed from a performance perspective Whatever's neat from a low-latency perspective is totally different from what we had previously when you know We would have to call the power company and the air conditioning company and get the lift in the air And you know could have taken three weeks just to get the equipment in there now It can be dialed up with a with provisioning software So Gary I want to ask you we talked we're talking earlier about some of the disruptive naval is actually the buzzwords are you know Low-latency big data analytics But virtualization is a technology that's been around for a while changing but relevant in the software-led infrastructure Talk about open stack in context to the data center And you share with the folks out that you look at the product road map for fusion I owe you get some visibility into kind of what's beyond just a hyperscale web But now enterprise hyperscale How does that impact the data center and and how do things like open stack and say virtualization? as virtualization is changing from hypervisor licensing VMware to a complete commodity and or software-centric world Yeah, I think all of these initiatives around virtualization Open stack are all helpful to customers to give them options. That's what's important And what we want to do is we want to make sure that customers can take those Those options in terms of how they want to architect their platforms, but do so in a way. That's simple easy Open software-centric So that they don't have to continually rework the infrastructure So when we look at the solution portfolio and what we're doing and what we will be doing It's a real focus on open Software-defined Enablement for people who want to choose the infrastructure that works for them You know some people like particular server vendor a some people like server vendor B Some people want to do their own thing or a fusion IO That's all goodness because we're enabling a platform that people can put in the systems of their choice You know we work closely with our OEM partners We work closely with our channel we're closely with our end-comer so we can deliver a folio of solutions flash memory solutions software solutions enhancements to open-source applications and then let The partners and customers build what's right for them. It's really important that we're not Mandating specific, you know hardware architectures. We're not mandating specific deployment models You have to deploy it one way or the other because the market is too big and there are too many Options for customers to say it's going to be one particular way Gary Talk a little bit about I mean one of the things you do want ISVs to do is is right to your SDK and take advantage of your Your software layer. Can you give us an update on where you're at? I know you're making a lot of progress with with my sequel per Kona Talk update us on that initiative and any other action that you got going there Sure, so the Software development kit efforts we found have been most successful when we help our customers integrate them into popular applications So as you mentioned, we've done a lot of work with my SQL Including groups like Kona also MariaDB To take some of the functionality in terms of the atomic rights as one example and get that into the the mainstream distributions It's other Applications that we're looking at as well. We think there's opportunities with data stores beyond my SQL You know, we're doing work with customers with Cassandra. We're doing work with MongoDB We're doing work with H base. So it's not limited to to my school, but we found that the best way is for us to work with You know partners to make that very accessible for customers in terms of application packages So that's been the primary focus where we're very enthusiastic about what's possible again Not only in terms of increasing the performance of these applications, which is important But also, you know getting longer life from the flash memory, which is also important Okay, so I have one question about Fusion so to how's David Flint doing? I know he was traveling around in the world lately I'm gonna chance to bump into him last month Dave and I and he was a little bit jet lag But he seems chill. He's very into the whole software month. What's on his mind these days? So I you know you can call him up and ask him directly, but he's doing well He does spend a great deal of time out with customers and and talking to folks to keep a read on the industry So maybe you caught him up here at the end of one of those trips But again as a company we find that this movement to to choice this movement to open solutions This movement to helping customers take the favorite servers that they've been buying for years and Transform those with flash memory and software into super servers. That's a model that's been working for us Working with our partners both our OEM partners and our reseller partners as well as our customers And I think across that spectrum of open software defined solutions for for folks to process more data and do so with less Infrastructure you're gonna continue to see more innovations from Fusion IO along those lines Excellent. All right Gary will listen. Thanks very much. We really appreciate you coming by the studio in Palo Alto Gary orange Dean senior vice president of products at Fusion IO always a great guest It's great seeing you again Thanks very much. Good to see you too Dave John appreciate the time. All right. Keep it right there Everybody will be right back after this commercial break