 Live from New York City, it's The Cube. Here is your host, Jeff Frick. Hey, welcome back. I'm Jeff Frick. We are here at The Cube's fifth birthday party at Big Date NYC in Manhattan. It's part of the Big Date a week. It's got Stratocon, Hadoop World, and of course, Big Date NYC. So now we're having our party, which is always good to have, and I'm joined here by my next guest, Gary McFadden, from Parity Research. Gary, welcome. Well, thank you very much. So last we saw you was actually at Big Date NYC 2013, so there was a lot of change in the year. Absolutely. Absolutely. I think the whole Hadoop thing is really taken off, and the thing that interests me the most about the show or the exhibitors of the show is that you can get a lot of data into Hadoop, but how do you get it out? How do you make it useful? What do you do with it when you get it out? Is it on-structured data? Is it structured data? Is it a combination? Is it schemalists? All of the above, right? All of the above, right, exactly. So I think really that's been, and actually I've been Jeff to all the shows since the beginning when it was just Hadoop World, when the Cube started back in, I think, 2010 or 2001. Yeah, it's our fifth birthday, right, so at least 2010. So since then, you've seen a progression of vendors coming in to provide services that actually enable Hadoop to do more than it does. So Hadoop started as kind of a badge-oriented type of solution that now, because of these other value-added solutions, can do real or near real-time processing. You can take the data out of it a lot more easily. You can use Hadoop, basically, as a repository, and a lot of the solutions out there are evolving to the point where you can basically make sense of the information. And I think that's really important. Right, right. It's data and information and information and insight, right? I mean, that's where we want to go with this thing. That's where we want to go. And then actionable business decisions made in real time, which we define as in time to do something about it, right? Right. So as the players, I mean, you've got the MapR guys, you've got the Action folks that just bought pervasive software, so they've got the predictive analytics piece sort of covered. Obviously that's Stonebreaker's old company, you know, a variant of Ingress. Right. You've got, obviously, IBM is a player in this space with their blue mix and their cloud capabilities and all of their information management pieces. Every major vendor has got a piece of, is part of the action, if you will, trying to build something on top of a dupe to make it more useful and make it more valuable. Yeah, the floor was filled with little companies, big companies, and everyone is certainly jumping in. So let me get your perspective, since you've been coming for a lot of years on this thing. Where are we on the journey? How, you know, I think we're past the POC stage, right? People are getting stuff into production deployments, but is it still early days, you know, the Giants are playing tonight, go Giants, are we first inning, third inning, seventh inning? Where are we? I think we're probably in the second or third inning. Second or third. I think we've got a ways to go. What's the next big hurdle to get us to the next inning? I think one of the problems is the storage issue, right? So you've got this issue of being able to scale out theoretically, exponentially, right? The nice thing about a dupe is if you need more space, you just add nodes, you add storage and whatnot, but what happens when you get too much information? You're into the petabyte, multiple petarite range now, and most of that data, you know, you're not going to access. You may access only 2% of it over time. I think there are a lot of figures around that, but actually a Wikibon article that I read recently is very interesting. One called Flate. Flate. Say it one more time, Flate. I just want to make sure everybody gets that, Flate, but I hadn't heard it until very soon. You sent it to me off camera. Right, it's F-L-A-P-E. It's a combination of flash and tape. And basically there's a great article on the Wikibon site by Wikibon's CTO, David Flore. And his premise is that at some point, relatively soon, as data grows exponentially into the multiple petabyte ranges and maybe even beyond, the thing that's going to get squeezed is the traditional HDD or hard disk, spinning disk, right? So tape has become much more resilient. Tape last has a meantime to failure of about 26 or 30 years versus disk, which is about five. And obviously flash is much, much faster. Right. Right? In some cases, I don't want to get into all the nuances of all the features. And we don't want the fees and the fees, but Flate is going to squeeze out disks in the middle of the area. I think so. So Flate offer customers is a much lower TCO for managing those huge petabyte scale environments and also accessing it at a relatively quick speed. So I think that's a piece that's interesting. The other part that's very interesting to me is the cognitive computing piece. So I was at the NoSQL event last month in San Jose and with that they had a cognitive computing component. And I think the idea of trying to get machines to think more like people, building neuromorphic chips to kind of mimic the way synapses or electricity in the brain works. How neurons fire and so forth is very interesting. And I think once you can get, Hadoop is the repository, you've got the data there but how do you make use of it? And I think that's the challenge that's going to be real paramount the next few years. Exciting days ahead. Well, Gary, thanks for taking a few minutes. We're at the fifth birthday party at theCUBE. We're at Big Data NYC. I'm Jeff Frick. We're on the ground. Thanks for watching.