 Live from New York City, it's theCUBE at Big Data NYC 2014. Brought to you by headline sponsor, Juan Disco, with support from EMC, Mark Logic and TerraData. Now, here is your host, Dave Vellante. Welcome back to Big Data NYC, everybody. It's really good to have you here today. We have been going wall to wall. Jeff Frick is with me today. Jeff Kelly has been here earlier. Yesterday and today, yesterday, we had our Capital Markets event. We had well over 100 people here for the Capital Markets event and celebrating five years of Hadoop World coverage for theCUBE. Brian Bukowski is here, our good friend. We just saw him at Oracle Open World. New setting here, of course, which is Hadoop World. Wow, two ends of the spectrum. Brian is the CTO of Aerospike. Brian, welcome back, good to have you again. Great to be here, Dave. So like I was saying, different ends of the spectrum, Oracle Open World, Oracle Red Stack, open in quotes. And now we're here at Hadoop World, totally different vibe and compare the two. Yeah, so the enterprise is here in force in a greater way than we've seen in other years. The ramp is there, but it's really feeling now the groundswell of these enterprises, these older school enterprises, the Fortune 50, finally really adopting the internet technology stack. And I gotta say, Aerospike being part of the advertising industry, being a part of retail online, it's great to see these guys really adopting. So the show moved from the Hilton to Javits, much bigger venue, does it feel smaller or is actually they filling it up? Yeah, I think they're doing pretty well. The last year here, it was a little small and it was definitely crowded. You're walking around between people and there, it's more open and it feels like about twice the traffic, we'll see. Yeah, okay, so from your standpoint, you're talking about the older school IT guys coming in, I hear the average age of Hadoop World is starting to trend toward me now. What does that mean for Aerospike? Well, it's great for Aerospike. So today we've announced our first Hadoop Aerospike integration. So what's great about this is since Aerospike is often used as a front edge database, so it's got your user profiles, it's got your session management, it's got recent activity that you're using to personalize. Having an immediate Hadoop integration means you can run Hadoop jobs over that live data as it's sitting there, without having to go out into some other system, to have to go out into some data lake and blah, blah, blah, you can just run your Hadoop jobs locally, or you've got your Hadoop system and you can just jam the results and use them online, not just for business intelligence, not for generating reports, but use it moment by moment directly out of Hadoop. I'm pretty excited about that. I mean, you live in the real-time world. I mean, you're talking about instantaneous repricing, repricing, I don't know what the number is, thousands, millions of times a day, I don't even know what the numbers are anymore. They even do it just on events. I see this happen, let's reprice. Well, give an example of that. Okay, so what they might do, for example, is sample Amazon and watch if Amazon is reprised. If Amazon is reprised, well, then I'm going to reprice, right? It's just like the two gas stations on the corner, the guys looking at each other's numbers and figuring out what to do. Great for consumers. Absolutely. I mean, a lot of people are concerned about all this. Oh, they're putting ads in front of me and they're trying to identify my profile and my preferences, but as a consumer, I mean, a lot of consumers would say, well, what's wrong with that? The competition from my time, my eyeballs, my preferences, isn't that a good thing? Well, you also have to look at the services you get. I mean, we get so many free services of such high quality from the behavioral work that Google does in reducing spam. I mean, they know about my conversations so they can reduce my spam. I moved from a paid service to Google mail just so I could get spam reduction. They do that with the behavioral information they have. So we did a survey recently and we asked organizations, are you shifting, substantially shifting resources from your traditional enterprise data world, enterprise data warehouse into big data? And an overwhelming majority said, we already have like 60% and another 35% said, we will by the end of 2014, 90 plus percent said, we're going to move. Now you're talking about all these older school IT guys, enterprise IT people here at the show, clearly that's a signal that they're shifting resources. So what's happening in your opinion? Are they sort of baselining their enterprise data warehouse, where are they putting their money? Are they spending as much as they were before? Or is it less, is it 30 cents in the dollar? Help us squint through all that. Sure, so at Air Spike what I see is we're seeing the companies that come in and they say, wait a minute, Amazon is entering into the credit card market. Google is becoming a credit card company. These guys are coming after my lunch and they're doing it with innovative services, innovative pricing, more customer engagement and they turn to their IT guys and say, well, look at all this money we're spending. I've got a million dollar contracts. I've got guys coming in at $1,000 an hour and Google is killing us on technology. We need to do what they're doing. Go buy some Google guys, go. And so the pressure is top down because those internet companies are just coming after all of the retailers and all of the enterprises that are out there. So of course they have to respond. The first thing they're doing is just saying, hey, I want the technology stack and get me some of those guys, get my feet wet. Retail is interesting, everybody that I know in retail has an Amazon war room trying to figure it out. They're also trying to figure out how to get leverage out of their in-store assets, their physical assets, trying to maybe tie those back into the online experience. What are you seeing there? Well, the funny thing was the news this week or the last week that Amazon's opening a store. Yeah, right, the physical store. So it can't be that bad an idea of a physical store, right? Well, I was talking to my kid, he's like, why are they doing that? I mean, nobody's going to buy stuff from the store. They're just going to go online. No, it's all about the brand. It's about the brand and it's about the showroom. So I'm still seeing a lot of experimentation and these guys are all very closed-lipped, right? So I've talked about it a couple of times in some of my meetups about how what Walmart has done in terms of Walmart Labs about trying to look for pricing signals, about trying to not discount was one of the stories I heard. These are all rumors because everything's under wraps, right? But if you have the power to discount and then know when not to use it, that sounds like a great idea. That's just, you know, that's meat on the bottom line. Walmart's interesting. You think about the drones and the same day delivery and things like that. And you think about Walmart. I mean, Walmart's got a lot of distribution centers a lot more than Amazon, you know? So everybody feels like, you know, saying, oh, you know, Amazon's just going to kill the world. Amazon has competitors, as you all know. You're helping a lot of them. And I just feel like the physical retail world still has a place to play. How can you participate in that experience? What is Aerospike seeing there? Well, so what we do at Aerospike is we find that there are a lot of really interesting next-generation projects that are happening. And that's everything from, you know, okay, there's only a couple of guys doing drone delivery, but there's plenty of people trying to optimize car fleets, right? And think about Google Shopping Express and what they're doing. They're using their internal Google technology, not Aerospike, but they're trying to get products to consumers through a whole new channel. And they need fast databases to do that. They need to be able to schedule all of that stuff that, you know, SAP talks about with HANA with, you know, fleets of trucks and optimization and stuff like that. Except these companies are using internet technologies, they're not necessarily in bed with the old ERP systems. So we see a lot of that kind of stuff where new projects are starting, whether it be customer service, in-store location, guy walks close to a Starbucks. Why should I wait in line? Really, I have to wait in line? I mean, I've got my Wi-Fi, I've got my phone. Why doesn't it know that I'm right here? Especially if you're a gold card guy, right? You get the same drink every time you're... Right, right, why doesn't it start off? My Starbucks app should know from the fact that I'm touching their Wi-Fi, it should pop up and say, well, do you want what you're usual? And then I just say, yes, and I stand over there. I mean, why hasn't that happened yet, right? So every retailer has all of these really interesting ideas. Do I know what's gonna work out yet? No, that's their business, not my business. But every single one of those needs real time, it needs personalization, that's the key to all of this stuff. And that's where Aerospike fits. Oh yeah, talk a little bit more about that because this is your wheelhouse. So what are the ingredients of that experience and generally, and specifically, where does Aerospike fit? So where Aerospike fits is on the front edge of that application. So instead of just using Hadoop to generate reports, figure out where to put a wind farm 10 years from now, which is a great example, right? It's a wonderful thing to know where the best place for that wind farm is, but then you've gotta go get the permits, it's gonna take 10 years. That's your ultimate batch job. So the... Of course. So on the other hand, moment by moment, right? That's where if you're coming close to a Starbucks, am I just driving past? Am I a regular customer? Am I a gold card customer? Do I wanna put my three most recent? Putting all of that insight online so it pops up right in your face in a sensible way, even sensing, how intrusive do you like it? All that kind of stuff. That's information that's coming out of your big data analytics, but then you need to react moment by moment, right? That's what people expect. I go into the app, what's there right now? You can't put Hadoop where, I take my iOS app and I call Hadoop. I mean, that's crazy talk, right? So you need a database in the front, that's your front edge database, and that's Aerospike. Aerospike is a great choice. We often see Cassandra, except we're 10 times faster. We've got use cases where I was using 30 Cassandra, now I'm using three Aerospike, wouldn't you rather do that? That means you can do more data, you can have faster data, you can even make more personalized applications. It really speaks a couple of things that came up in the panel yesterday, where Avi Mehta was talking about, everyone's got data, but do you have differentiated data? Do you have special data about your customers that only you have, and how can you react and really leverage that? And then the other thing we love to talk about all the time is real time. Well, what is real time? Real time is in time to do something about it, right? So do you have time to catch that guy walking down the street to offer him his latte, pop on, and we've already deducted it from your card. So it's real time relative to what you're trying to accomplish and take care of something. Exactly, and I think the mobile economy really does trigger a lot of real time use cases. The internet did it already because I come to your website, I've got that attention for a moment. I have to present an offer, decide to discount, give a specialized recommendation right then. That's been our wheelhouse in terms of advertising, in terms of retail. That's where you put in an Aerospike server. You take your Hadoop job, you pour the result into Aerospike, but Aerospike's right there moment by moment. Take that over into mobile where you're driving past. You've got traffic data and things of that nature. You're trying to, you know, all of those things are true real time use cases and the ages come to take big data, move it into real time, everyone's talking about it. That front edge database is Aerospike. Right, because you can build the applications. It's a Hadoop application. Applications. It's not a Hadoop, it'll be cared to. It's applications that deliver value. And that's why we put so much energy into today's release of our Hadoop connector. So being able to take that Hadoop insight and having pre-built code that allows you to jam the result into Aerospike and use it right now, we think that's a very compelling decision. Technically what is a Hadoop connector? What does that mean? Technically it is an input format and output format. It sits at the, basically at the Hadoop layer. It's above HDFS, which gives it some extra speed so you don't have to go through extra layers. And essentially you can just emit data or emit results of a Hadoop job directly into Aerospike or use Aerospike with the input formatter and use queries, use indexes. You can do things like run a Hadoop job just over people in the last five minutes because Aerospike is not just key value but indexed. So it's all on our website. It's under our lab system. You can go in there and check out the source code. We've got a bunch of examples and session management, et cetera. Well, that's cool. You mentioned Google a couple of times. Does Aerospike, do you get inspiration from Google? I mean, in some respects what you're trying to do is bring a Google-like capability to the masses, Google's competitors, in any case, is the Google protection unit. Well, we do. And our original customers in advertising, all of them were saying, we want to do something innovative in advertising that Google perhaps hasn't done. For example, we're here in New York, our longest and oldest customer app, Nexus, of use of Aerospike. Their idea was they said, well, Google is running this black box where advertisers come in one side. Let's open it up. Let's create a trading system. This is the home of capital markets. Let's build an advertising trading system. And that's been an amazing success for them. The first billion dollar market cap company, start-up out of New York. Brian O'Kelly and all those guys have done an amazing job. Love them to death. But they said, hey, look, Google's a closed system. We want to do an open system. We want to have auctions for every single ad on the internet. Now doing that at three million times a second faster than anything's trading in the capital markets. Great success story. So talk a little bit more about how they use Aerospike. I'm interested in that case study. Sure, so you need a user data store in order to determine the price of an advertisement. So you've got a person, and you've got a piece of content, a webpage. Those two are coming together. You have an opportunity to deliver a message. You've got 100 companies. They literally have 100 companies bidding on each one of these ad impressions, every ad on the internet that's going through AppNexus. So the problem in the old days, so premium content, it's easy to price. If it's on the New York Times, those guys fill up their ads. There's not a problem. But what if I have catster.com? They're kind of in the long tail. And so how do you price a given ad on catster while you have to create an auction? That's the best and fairest way to determine the price of something in the long tail. Well, how do you run an auction? You can only do it if you know a lot about the user. What's their current behavior? You don't know the user per se, right? You don't have a phone number. All you really have is a cookie and perhaps some recent accesses, recent search terms, things like that. So you've got these 100 companies out there. The first thing they need to do is they need to look at the user data. That's an aerospike. So they do that first database lookup. It has to happen in less than a millisecond because this auction, I mean, it's right now, right? There's this ad opportunity that's gonna get placed and it's gonna get placed in 100 milliseconds, but the ad has to get there. So then you've got, you've only got about 10 milliseconds to do a real time price of essentially this capital market asset advertising to this person right now. So all those 100 companies, they take that user data out of aerospike, they run their own proprietary algorithms on it and they price it. And then the guy with the highest price wins. And so when they initiated that project was from a blank sheet of paper, it sounds like it was, it wasn't like they were sort of enhancing an existing system. This didn't exist before, is that correct? Well, it was certainly Brian and the AppNexus founders, it was their vision to create this marketplace. When they originally launched, they launched, I believe they'd had an in-memory system that was of their own devising. They're all wonderful technologists over there. They switched over to a system called Schooner that was bought by Sandisk, but it wasn't clustered. And that was the big problem in their system. They knew they needed Flash and SSDs, Schooner was great at that, but wasn't so good at clustering. So as these guys blew up and they just needed more and more machines and they needed to add machines every month, they needed a clustered software-based solution. In order to handle, yeah, the week-by-week scaling that they were doing and that's when they picked up Aerospike. What's happening in Silicon Valley these days, Brian? John Furrier's not here, he's out doing some stuff with the crowd chat team this week, but give us the skinny on the Valley, what's the buzz going on? Well, I think that the big conversation is how big is the bubble and are we really that close to bursting? That's what happens in all the conversations I have. So, you know, I was talking to a buddy of mine, an old Fusion IO guy, in fact, and he's getting his company funded and he's like, well, I really want to get it funded soon because it's something going to happen. So I think there's, I'm starting to feel this nervousness that we've been talking about a bubble for a year and that's usually when the worm starts turning and something happens. So I'm starting to feel a little nervous. I hope it's not true. This week's been Ebola, the economy, Europe. We've got Ebola, we've got a stock market drop. They're talking about it on the way back to the airport and the taxis are offering you stock tips and you know we're close, that's what happened. Yeah, good, that's a great point for that one. The market's having a good day today, maybe it's a good sign on Friday, you know, buy low sell high, but boy, it takes guts to buy into the October markets, doesn't it? So that's interesting guys, because it's been bubble-licious for a while now. I mean, really essentially coming out of the downturn, you know, 2010 to now has been really pretty frothy and you know, you saw the Cloudera, we just had Mike Olson on and I don't know how much it was, 750 million, I guess, from Intel. I mean, pretty crazy. It's interesting, you know, when the market goes up, valuations of private companies go up when the market goes down for a while, anyway VCs try to keep the valuations up for those. I don't know, I mean, do you think maybe this is just a sort of healthy respite? Well, I hope it is, I mean that's what you always hope is that instead of going bigger, bigger, bigger pop, that you know, you get a little flattening out, you get a little rest from it, so one hopes that that's certainly what's going on. I mean, the first thing though is that, I think we're seeing the beginning of the end of the big mobile boom, right? So mobile drove so much of what happened in Silicon Valley in the last three years and it's still going on, right? Uber is out there and you know, we saw, was it Halo, is that how you pronounce it? We saw Halo drop out of the New York taxi scene just this week and give up and just, you know, abandon the US but we're starting to see consolidation in mobile. The App Store has 100,000 items in it, you can't really break through anymore so it's really about, we're in the market share portion of mobile, people aren't out there establishing new mobile companies, it's the fight to actually get to mobile. We're starting to see the biggest phones these days in mobile are actually lower spec than the ones a year ago. That's what happened with the Galaxy line is the next one up actually has lower resolution and a slower clock speed. So I feel like the mobile, the big fuel of mobile is starting to flatten out a little bit and hopefully we just take that in stride, big data is coming on, there's the next big things about to happen and there's no pop here. Well maybe that means the mobile infrastructure is now in place and it's going to get stable, I mean apps are getting much, much better. Absolutely. And so okay, now we can maybe build upon, we talk about, we've been talking all week about riding on top of that digital fabric and the digital fabric is getting built out and it seems like this, that's where a lot of the innovation is occurring. In fact, the way people are applying big data is sort of the interesting theme these days and a lot of your customers, right? Well absolutely, so with mobile now becoming much more robust as you say, people okay, I have my app, what do I do next? How do I get a mobile alert on there? How do I do real time pricing on my mobile? How do I use all of the data that's exhaust from my mobile systems since mobile has got more engagement, more page view by what about a factor of four than your PC, you're out and about with it. Now that that's seeped in and people are just used to it, right? We've got more big data opportunities. So that's definitely what we see, more real time pricing, more air spike instances, tracking mobile apps and enhancing mobile apps. So what's next for you guys? We saw you at Open World, here at the Duke World, I'm sure you're super busy end of the year trying to get a jams of business into Q4. What's next for you guys? So we've been having great luck adding more clients and reaching out to more developers. I gotta say Go is white hot for us right now. Really? White hot. So I don't know how much you guys see Go around here. I would say that's another thing in Silicon Valley that has not yet really reached out to, I would say the New York and some of the other tech scenes. Air Spike was at the dot go conference in Paris last week. We saw a huge positive response there. Go is a very interesting next generation language. It inherently uses multicore. It inherently is sort of an expansive multi-cloud system. It has the simplicity and ease of programming of Python. It's a very comfortable language. And yet it has a lot more of the parallelism of say a next generation Java system without a lot of the verboseness. So it's a super fast language. It needs a super fast database. That's why people are reaching out to Air Spike, often for sysadmin uses. But we're really seeing Go coming on and we're very bullish about it. A lot of developer buzz around Go and the Docker of course is the other. Absolutely. Thank you. All right, Brian. Hey, thanks very much for coming on theCUBE. Always a pleasure, some good sound bites as usual and appreciate you coming on. All right, great, thanks for being here. Keep it right there. Jeff and I will be back with our next guest. This is theCUBE. We're live at Big Data NYC, we're right back.