 Welcome back, I'm John Furrier, you're the founder of SiliconAngle.com, SiliconAngle.tv, and we are live in San Francisco, California for the Node Summit conference. This is a conference that celebrates the rapid rise of Node.js, a development environment that is just taking the world by storm. All the top alpha developers are here, hackers are here, executives are here, entrepreneurs are here, venture capitalists are here, and this Node.js product is really becoming a framework for expanded productivity for developers, bringing out new mobile solutions to the marketplace, and SiliconAngle.com is covering with theCUBE. theCUBE is our flagship telecast where we go out to the events and talk to people and interview them and get the knowledge and share that with you, and we're here talking about what the impact is from developers building new mobile apps, web apps to market, getting funding, and all this activity around the developer marketplace that's driving massive innovation. Obviously, San Francisco's Workground Zero is for Node.js, and we're here broadcasting live with SiliconAngle.com, so all the coverage you want to see go to SiliconAngle.com every day, and that's where the innovation reference point is. I'm joined here for this segment with David Floyer, who's co-founder of Wikibon.org, SiliconAngle's research group that does all the deep dives on technology, goes in the weeds, talks about all the key aspects of tech. David is a guru in I.O., in storage, in systems, so David, welcome back to theCUBE, and great to have you. Oh, thanks, it's great to be back again. So David, we want to talk to you about what's going on here at Node.js. Obviously, I want to get your perspective. You just wrote a manifesto on Wikibon.org, and we wrote a blog post on it at SiliconAngle.com with all the links there, so go to SiliconAngle.com and look for that post if you're interested in the manifesto, but you really did some deep dives around the future of systems around I.O., and before we go into depth on the manifesto, I want to get your perspective around what's changing in the architecture of the internet and the internet applications that drive that. Obviously, the systems that drive everything have storage, they have compute, they have I.O., network transport, and now applications on the iPhone or mobile devices or the web need to leverage that environment, and we're seeing that in two massive surges around cloud computing and the rise of the computer at the edge, the mobile device, the web app. So what's your take around Node.js, this Node Summit event, and I.O. in particular? Oh, that's a great question, and it's great to be here at the summit. There's a buzz, there's a huge number of young, extremely talented programmers here, and they're tackling the problems of this mobile computing, they're tackling the problems of vast amounts of messages going between people from machine to machine, their mobile messages, and what they're doing is providing a framework where very high speed transport of these messages, analysis of these messages, usage of these messages is going on, and the speed is more important than absolute certainty of delivery of the message. So it's a new paradigm, and obviously that's going to be a culture shock as it goes into the enterprise, which it will. It'll be a culture shock against people who are used to guarantee deliveries, absolute certainty, acid properties of database to be dealing with environments where guarantee is not absolute, acidity is not absolute, and they will have to find new ways of solving these problems. So it's a very exciting time. So let's just talk about the applications on the web. So the web we're all used to Facebook and chat and instant messaging, but we now have moved from this PC era to an era of mobility where you can have a cell phone, be logged on at home, so all these things are going on and massive amounts of people can connect to any different application at any different time. It puts a new kind of, new constraints on subsystems or these computing systems. So describe that a little bit and what that means, and the challenges of a developer who's just writing code, JavaScript, or building a game, all this stuff has to take into account, all this complexity. You have to abstract it out as much as possible and rely on great standards like HTML5. You have to extract as much as possible out into platforms like Node.js and make it quick and easy and simple to develop these applications with less skilled people. And that's what the joy of Node.js is, is that it's so simple, so quick and easy to develop. It's a set of nodes being spun up all over the place and interfacing with each other and relying on services from other parts of the infrastructure to do the delivery to the Android or the Apple or do the delivery across on a global scale or planetary scale across the whole network. We get a lot of viewers from just in debt TV and they're familiar with gaming and gaming has a lot of concurrency, a lot of simultaneous users, a lot of in-game complexity. They use a lot of big data to track all this stuff. So there's a lot of stuff that goes onto the covers that a normal designer or programmer might not have to take advantage of like operating systems, things like threading, all this complexity. So that's kind of new to this world and that's what this conference is about. You have been doing some real seminal work in this same area relative to big companies. And we all know about Apple and Facebook have massive amounts of constraints and it has a lot to do with the storage and IO. So let's talk about your manifesto, your research piece that you posted called the IO infrastructure, IO centric infrastructure. A little bit further, what was your key findings in your research? Well, the key finding, and I've been talking about it for some time, is that the disk systems themselves are the constraint on so many applications. If you're going to make sure you have guaranteed delivery of data, you have to put it onto some sort of disk. That's the only way you can guarantee that you'll find it when the system goes down. So that is, that constraint of disk is that disk is just so, so slow and very, very narrow path to a huge amount of data. And they have not been speeding up at all. The amount of data you can store is massive on disk. It's very cheap per gigabyte, but the cost of it per IO has not come down at the same rate. So what's interested me is the use of flash devices. And one of the things that happened very recently, which was seminal, was the demonstration of a billion IOPS system. It just took eight of these processes with 64 cards to deliver a billion IOPS. That was between Fusion IO and HP in San Francisco just a few weeks ago. And what I did in this one is... Hold on, just back up a second. So HP and Fusion IO, so HP servers. HP servers. And Fusion IO. Proline 370 servers and Fusion IO, the IO memory cards, memory to Duo cards. Which is SSD or flash, right? Yeah, which is the SSD, solid state, MLC cards. So an amazing amount of density of IO that could be generated. Now, very small IOs, 64 bytes only in size. So a very trivial in size compared with most applications, but really an amazing achievement. So can we go to your slide on the manifesto and talk about some of the components around this IO centric architecture? Right, so if we look at the slide, you'll see that there are five layers in this IO centric infrastructure. So the top layer is the working flash storage layer. So what we're looking at there is what came out of this demonstration is a very, very tight connection between the processor and the flash itself. So you're doing something called atomic writes. And what that means is that instead of going through the IO stack, which is thousands of instructions long, you're doing a single instruction and writing it to, in one pass, to the flash. And that is orders of magnitude faster than the previous ways of doing it. So that very tight coupling, that use of atomic write, the first demonstration of this is really a breakthrough. And that allows huge amounts of IO to be processed very, very quickly indeed. And that's got real ramifications which we'll come to a little bit later on how you design systems in this sort of environment. So that's the first layer. The second layer is that, if you take the third layer next, the one in the middle, that's series of shared infrastructure, shared flashes. And lots of flash only or mainly flash devices which are connected to that first layer. And then between the two is an active management layer to manage the flow of data from the top to that shared layer and back up again. This is the active data that you're focusing on. And the thesis is that almost all active data will be in flash over the next decade. It'll be where active data lies. So that's the first three of those. So who's impacted by this? Obviously, there's some real success stories around here. Talk about the companies that are affected by this. I mean, also we've been following the rapid success of Fusion I.O. when public. They've been on theCUBE since our first Cube gig. Solid fire. And then EMC, the big whale in storage who's actually been servicing that market of normal storage. So take us through the horses on the track here. Well, you've got Fusion I.O. Obviously a relatively newcomer. They came on board four years ago using PCIE cards as opposed to SSDs. You've got HP who've been a partner with Fusion I.O. on the ProLiant servers. And they've embraced this technology. So those are two leading horses. And HP is trying to bring together servers and storage into a single unit, into a single component that you buy together. And that's a very exciting strategy for them. It makes a lot of sense to bring those two things together. So the systems expertise of HP is going to be very important in looking for solutions in this area. You mentioned EMC as another player in this area. They have project lightning, which is their announcement of getting into the flash and I.O. area. That's very interesting indeed because what they're looking to do is to put PCIE cards into the servers and also manage those together with the shared storage arrays in layer three on the diagram. So they have FAST, which is part of that active management of data level. They have FAST. They obviously will have caching to begin with. But the really interesting thing is when they get further down their program and start to introduce cash coherency across the servers and the layer three, that's very interesting technology that will be necessary. Let's talk about the opportunity for Fusion I.O. Solid Fire and say EMC, for example. So here at Node Summit, talk is developers on the front end flexing their muscle becoming more back end. Like we talked with Matt Rainey, who's the founder of Voxer, one of the most successful fastest growing apps. And we chatted last night at dinner over a million, I think a million users a day or a month or whatever, large number doing a lot of I.O. stuff. I asked him specifically if he's using any innovation on hardware, he said, I don't know, we just go with the hosting. So for him, they're kind of ignorant to what goes on in the back end. That's an opportunity for EMCs of the world. So how does EMC become better for this market as the market is obviously rising the tide on the developer side? How does the traditional storage vendors and the newbies like Solid Fire and Fusion I.O. How do they service that market? It's a very, very interesting discussion and being part of that discussion obviously, the I.O. that he's talking about is audio, it's large amounts, if you lose a bit or two, it's not the end of the world. So it fits into the paradigm here at Node very well indeed. And for that particular application that may or may not be the right usage of this technology. But what is important is when you go into the enterprise, guarantee delivery becomes much more important. There is a desire to reduce risk, the impact of risk if you lose a transaction or something of that sort. So in that case, the guarantees come from being able to write as soon as possible to a persistent medium, in this case, the flash device. And that is extremely important for enterprise type applications. So EMC, if they go into this business, they're going to use their strong relationship with VMware, being able to add technology to the VMware level, the hypervisor level, being able to improve the I.O., being able to write that I.O. and guaranteed a right security of that I.O. Those are areas which EMC can really add a tremendous amount to the infrastructure. So you're saying, if I read you correctly, the Node Summit, the Node.js opportunity actually helps these guys. EMC and solid-firing. Oh sure, yes, yes. There is a market out there that will want to do lots and lots of I.O., and that I.O. on disk is just as slow on a Node.js system as it is anywhere else. And these technologies will come in, in my view, up and down the stack. Let's talk about HP. HP storage and networking is doing very well within HP. Obviously, you mentioned the I.O.P.s, one billion I.O.P.s with Fusion I.O., that was a demonstration with HP servers. Obviously, HP storage has a huge acquisition last year with 3PAR, and that's large-scale storage. So what does that mean for 3PAR? What does this all mean for 3PAR? Can you elaborate on the 3PAR, HP 3PAR storage component most of the I.O.? One of the announcements that HP have made is the bringing together of servers and storage in packages, delivering them together. That's a very interesting development. The other very, very interesting development in this mid-range and this management across these devices are the federated storage that 3PAR has announced. So the ability to be able to move applications in the same way as VM, but do it from array to array is very, very exciting technology to be able to, if you like, virtualize movement across those sub parts of it. So there's a lot of technology there that they can bring to bear to help in this new type of a large I.O. environment. Obviously, they're going to have to make investments in the flash end as well within the array side, but they have a lot of interesting technologies to add that will be contributing to this system. So this category for hurricane that I called earlier, called node.js, is really on a collision course with these big guys like HP and EMC because we heard from Theo Schlossschnegel, who runs OmniTI, operating system guy works on large scale, big deployments, a lot of enterprise, a lot of service providers. His comment was legitimate around that a lot of these guys who are kind of getting more back-end capabilities with node just aren't visible to all the problems that go on in the lower level system components. So people who understand systems programming and systems design have been there before. There's more operational processes in place and it's always up and running. So these guys are going into an environment they just know nothing about really and they're kind of ignorant. I mean, they don't, they could know in theory, but at practical purposes, this is really a perfect storm for HP 3-par and EMC Lightning if they play their cards right. Is that correct? Absolutely, yeah, because at the end of the day you need operational systems that stay up and keep up and that needs solid processes and solid quality processes, all of the normal infrastructure methods that have been done. At the same time, the Node.js people are doing things in record time and developing things and they have their massive contribution and the two together is going to be very exciting. You know, we've had Fusion I.O. on theCUBE and we've had SolidFire, great technologies, but SolidFire in particular, the CEO there used to be with Rackspace. So he understands Rackspace and we just heard from Theo that Rackspace has been quietly doing a lot of Node for their deployments. So it's interesting to see how SolidFire is very well positioned for this as to keep up with the scale of provisioning and deployment of I.O. and storage. And SolidFire, they've got a very interesting angle on particularly important in this area, which is that the key cost item in Flash is writes and those are the things you've got to monitor and allocate out and cost for. So they've thought through very carefully if they cost allocation infrastructure, there's going to be allow service providers, for example, to be able to throttle or give specific amounts of I.O. And that's what people are going to be buying. It's not, there's more than enough gigabytes around the place. What they'll be buying is the I.O. capability and the rapid I.O. capability and you've got to pay for that and they don't want to give away I.O. If they're spare I.O. They certainly don't want to give it away. Let me ask you about another company called Data Direct Networks or DDN. So I've been having conversations with John Luke Shadlane who came from HP as a senior executive now at DDN. He ran a lot of the information and governance services side of HP for years. Brilliant man, he's been on theCUBE. I got some great interviews with him and one-on-one with me. But fantastic technical and business leader. I've talked to their technical teams. They've come out with a very successful approach around object store where you can store all this stuff. So where do they fit into this equation? Obviously that approach seems to work here. What's your take on DDN relative to all these new trends? Well, what we didn't go through is the last two layers of the manifesto which is the archive management layer and then the base layer which is mainly disk based. I mean, disk is not going away. Disk has got a very important part to play. So what you want to do then is that within the whole process, when you gather the data at the beginning, you're going to be doing the metadata and the indices right from the get go and be putting that into the third layer. And the advantage of that then is that when you create the archives or when you create the long-term backups, you've got all the metadata there. You've done the deduplication. You've done all the stuff to condense it down. And then at the back end, what you need is very high-speed devices, very cost-effective devices, both to store it and to distribute it across a network. So WAS is a great product that DDN have. Web object systems. Yes, web object systems. It's an object-based storage mechanism which will allow you to distribute stuff across the whole of the network. And DDN have a great set of very high-speed low-costs. So that's a cloud opportunity and multi-geography opportunity, right? Absolutely. And again, this emphasizes, if you want to get that data back from that lowest level, it emphasizes absolute top speed of getting it back, especially sequential and large objects such as video. Okay.