 Live from New York, it's theCUBE, covering Big Data NYC 2015. Brought to you by Hortonworks, IBM, EMC, and Pivotal. Now your host, Dave Vellante, and George Gilbert. Welcome back to New York City, everybody. This is theCUBE. We're right down the street from the Javits Center where Strata Hadoop World is going on. This is our event within the event, Big Data NYC. theCUBE goes out to the events. We extract the signal from the noise. Peter Goldmacher is here as the CMO of Aerospike, friend of theCUBE, friend of SiliconANGLE Wikibon. It's great to see you again. Last year, just about a year ago, we were up, we had that great panel that you participated in with Amy O'Connor and Abhimeta, and Sparks were flying. And now you're back in the technology world with Aerospike, a company that we know pretty well we've been tracking for a while. So how's that going? It's going great. As you can imagine, startup life is dynamic. Last time we talked, I was running marketing, now I'm not. So I joined eight months ago as head of strategy. We had a new CEO. I came in with him. We changed the organization around a little bit and now we've brought in a proper head of marketing. So I'm back to just the strategy role, which better for me. But it's great. You know, I was on research, I was on Wall Street as a research analyst for a long time, like George. And it's a great opportunity to see the world and taking that knowledge and coming back into industry. I cut my teeth at Oracle in the 90s. Coming back into industry, especially the company like Aerospike that's so unique and so differentiated in what we do. And try and get the word out and try and make it a commercially successful business. Certainly is a challenge, but it's an enjoyable challenge. So let's talk a little bit about the uniqueness. I'd like to do a couple of things. Aerospike, the uniqueness of Aerospike, where you guys are winning, some of the use cases that you're taking down, how you're different from the other guys. And if we have time, then I want to get your perspectives on the rest of the industry. There's a lot of interesting stuff going on. But let's start with Aerospike. Yeah, so one of the things we didn't have when John Dillon got there and when I got there two weeks after John was we didn't really have an identity. We were just another NoSQL database and we were going to market like all the other NoSQL databases. And pretty quickly we looked at our customer base, which at that time was about 85 customers and said, what is it all these guys have in common? And what they had in common is they needed to take an extremely large data set and run pretty complicated rules engines on top of that data set to get very high quality answers based on a large set of data. So you know our first success was in ad tech. So when you get a banner ad that's strangely relevant, what happens is every morning a Hadoop profile is loaded into Aerospike and then as you're online, all your activity on the internet goes into Aerospike and they look for a match between, well we know Dave and this is what Dave is doing today. What kind of ad could we serve him that has a higher probability of a click through? So the ad tech guys write the algorithms but the database is the critical component in getting all that data live and actionable so they can make it an immediate decision. So analytics are great if you want to know what's happened but what we do is we make sure you can look at everything that's happened and say, okay I can write an algorithm that says make this decision when I see this thing and so Aerospike as a database underneath that enables all that. Yeah and we were talking about it. Obviously ad tech is one use case, the other is fraud detection. There's a big difference between the good and the not so good. You see that in ad tech, we talked about fraud detection, we'll talk about that in a minute. But so I'm presuming you guys are behind the good because that's what you're- Only good. That's what you're about. We only do good. Right, I mean you're not about the, you were talking, you know, the sort of low cost, you know, sort of reduce your investment in database infrastructure. That's not your gig, right? That's not the gig we go to market with but that factually, if you understand the internal workings of how we do what we do, there is an enormous cost saving component but that's sort of the secondary message because the primary message is so powerful and the primary message is we enable you to do things you could never do before because our database is fast enough to support decision engines and rules algorithms that you could never process enough data in enough time to make a decision. So let me just maybe an example helps make the point. Why is it that when you get on a plane in Boston and you land in Paris and you swipe your credit card for a cup of coffee at Charles de Gaulle Airport, it gets denied? Well, all that data exists. I know who you are. I know you bought a plane ticket to Paris because I've got the credit card transaction. I know you're staying at the Georgetown because you booked the hotel. I know you had a bowl of chowder at Legal Seafood eight hours ago at Logan. So when you land, all the data exists. I know it's you. It is knowable that it's you. Why can't I approve that transaction? You can't approve the transaction because in the old world, you can only write a rules algorithm that's sophisticated enough to handle the amount of data that you can process in a very tight time window. For credit cards, it's usually about half a second to three quarters of a second because the database is slow. The reason writes to the database to populate the rules engine is slow. So if I can guarantee you that same SLA of a half a second or three quarters of a second and the opportunity to look at much more data and write a much more complicated rules engine, instead of just saying, okay, this is Dave's credit card, is he current on his payment? Is he over his credit limit? Is he in Bulgaria? Very simple questions. I can now ask all of those things and I can also look at all the other data. I know, what charges has he charged that might correlate this transaction? Is he using a device that's familiar to me? Is he using a vendor that's familiar to me and a known risk? Is he using an IP address? So all of a sudden, the opportunity to look at all that data, you get much better scores. And for credit card risk and fraud, this is 100% margin stuff because the biggest problem in the credit card business is not fraud, it's false negatives. Because if you're doing a transaction and it's valid and I deny it, that's money off the bottom line. And so one other very common trait of all our use cases is it's all mission critical. All of this stuff directly impacts revenue. It's not content management system or a product catalog. It is revenue impacting stuff. Yeah, so I love the story and you're right. I mean, the database business is so crowded. If you want to go to market with yet another no SQL database, you just get lost in the crowd. That's great going over there and yapping about it. But everything you just said is about changing business outcomes and driving value. Speed of scale. So how's it resonating with customers, particularly in the financial services industry? I mean, I was giving you my example of some of my frustrations. Because fraud detection is so much better now than it was 10 years ago where you had to wait six months and maybe you'd find out because you were sampling data. Okay, that's great. Now that we're in that semi real time zone, I'm really not satisfied. Can you make a dent in that? Yeah, so one of the things that's happened over the last eight months since I've joined is we've completely changed the sales and marketing team and the go to market strategy. And we now realize there is an enormous opportunity in a couple of verticals. Ad tech, which we dominate. Financial services and telecom. So we are going after primarily those three verticals. And what's interesting is if you look at the Mongos of the world and the data stacks and the couch of the world, these guys have a completely different game. They're an Oracle alternative. They're saying, hey, look, the product, the database is no longer a product, it's a category. And if you are doing something that's not ideal for Oracle, is there a better, cheaper way to do it? If you're building a product catalog, content management, you know, Mongo is a great solution for that. If you have a large amount of data, you're trying to get written really, really quickly, Cassandra's fine. But if you want to take in a ton of data and you want to have access to that data really, really quickly, we're really the only solution on the market. And that's what we specialize in. High performance, no sequel. Great. Well, so tell us like, you give us a great example for financial services in ad tech. So let's assume there's that same budget of X, you know, a couple hundred milliseconds. What can you do better that, you know, when you can reduce, well, if the database has the same budget in terms of within that, say 700 milliseconds, how much better can you do the targeting, you know, the pro, what's the profile that you know better? I'm glad you asked that question. So the SLA in the fraud world is half a second to three quarters of a second. The SLA and ad tech is 10 milliseconds to 100 milliseconds. So it's- For just the database part. For the whole thing. The whole thing. I mean, you got to get that out of, because you got to bid for the ad, you got to serve the ad, and there's a lot of things that happen. So you're asking, how do we add value in that process? We don't write the algorithms. So we've got customers like XLA and Atmexus and Tapad and, you know, big, big concentration of customers and right here in Manhattan. And their secret sauce is the algorithms they write. So that's a great question for them. But when they come to us, they know that whatever they write, whatever correlations they write, whatever rules they write, we can get them enough data to populate that engine and get a response in that extremely tight SLA. So remember, we're only a database, but we enable this next generation of applications that embeds analytics, but correlates it against what's happening right now. So Dave, you and I have talked about this whole Gartner discussion of hybrid transaction analytical processing. To us, that's crazy talk, right? Because if you have one database that does both things, that's fine if your SLA is like four days. Our SLA is again, sometimes microseconds, frequently milliseconds. Yeah, that's an estimate to you guys. So it just, that's just not the world we plan. That's just not... The TapPad guys, I think we're telling them, I don't remember the exact data, but he was telling me that they essentially apply that algorithm and reprice their ad by, like literally thousands of times a day. It's a market. Versus, you know, some of the competitors which is once or twice a day, maybe if they can do that, because this human's doing it, you know? And so that's pretty, pretty stark. And now what about telecom? Is that customer churned stuff or is it more? So I'll give you a telecom example. Telecom, I think we all, fraud and telecom we all relate to. So we have a customer, Alcatel Lucent. And Alcatel Lucent is actually a reseller billing systems to all the carriers, AT&T, Verizon. So all of us have had the experience where you get a text message that says, hey George, you're about to go over your data limit on your plan. Do you want to buy a gig for 15 bucks? We've all gotten those texts. That's because they can't be more specific because they have their usage in one application and they have their customer billing and customer profile in another application, right? So they say, if anyone's going over a limit, doesn't matter who he is or what the plan is, sell him a gig for 15 bucks. But if you pay 15 bucks for a gig and then your billing cycle ends a day later and you use 25 meg, your good and warm feelings that you usually have towards AT&T are tarnished a little bit, right? So they've written a new application on AeroSpike that says, look, let's correlate data usage on the billing plan. And if you remember, the old billing plans were on voice minutes. Now it's all about data and the amount of data that we're charting and the dataset has exploded, right? So you can't, the old way just doesn't work anymore. So this application that I think is live in a Vodafone Italy is basically if you are about to go over your limit, they know what your usage patterns are. They know what plan you're on. And they say, hey, George, you're about to go over. I know you've got two days left in your plan. I know you're gonna use about 50 gigs. Do you wanna buy 50 meg? Do you wanna buy 100 meg for two bucks? So you think, well, why would they sell less when they could sell more? The whole point is if they can reduce churn, if I can reduce George's churn on a $200 a month plan by one month, that trumps any $15 a gig that is gonna cause him to churn early. So what we love about that story is, this is a really different way to think about things, right? You have an opportunity to do something you've never done before, because you're thinking, boy, I could write a completely different kind of application if I could get all this data and process it really, really quickly. So we love that. So how about that? Telco's deciding it's good business not to screw the customer. You know, the examples you're giving us are very quantifiable business outcomes. And it's hard at infrastructure software to sort of sell on business outcomes, I should say to price. We also have this sort of thesis that we're going through this slow motion price collapse and infrastructure software. Because even if you're not open source, you're competing against open source. So how do you live in that world? And for those who, you know, say they make money by helping customers run the software, how does that work? So there are a lot of questions in there. So let me touch on a couple of topics. One is open source is a problem. Every single open source company will tell you their biggest competitor is themselves, the free version of the product they're selling. And if you look at kind of the way open source got spun up is you had total abuse at the hands of Oracle and EMC and pick your favorite large enterprise vendor. And they had just hammered their customers because there were no alternatives in the market. And then you had Linux, right? And Linux was unbelievable because it was this product put together by a bunch of hackers that became commercially successful. So people said, oh, open source, I like that. But open source as it exists today has almost nothing in common with Linux. Linux was a bunch of like-minded guys, probably a bunch of commies that got together and said we don't care about money, we just want to build great product. That is not the ethos of any open source company in the market today that's taking venture money, right? Guys think, hey look, if we build it and they come and I can get distribution, then I can figure out my monetization model later. So look at what we know about Mongo and Datastax. These are enormous distribution mechanisms. The bulk of the money invested is in distribution, not in product, right? So they're saying if I can get to the market and I can be big and I can be standard, then I'll figure out monetization later. They have told, I mean, like I think I've talked to, if I remember, that senior VP of sales like at Mongo and his attitude was, yeah, we're going to have to build a real sales force to call on the senior people just like Oracle does. Yeah. Yeah, but then you wonder, I guess it's not clear how do they monetize? Although they say they spend a good chunk of their R&D on manageability. So I guess our question is, how can everyone, every product help customers manage, but then who's managing all that manageability? It's just 27,000 different consoles. Yeah, but that's a problem that's as old as software from more than one vendor. Yes. So we're not treading any new ground there. I do want to make one point, and if I can just kind of revert to my Wall Street days, the thing I always worry about is what happens if funding dries up for open source companies? And these guys have invested all their money in distribution because that's what they've identified as their key differentiator. They don't get distribution. They don't get breakout velocity. The money dries up, and they've got this enormous sales force, and they've had a revenue model that's, the nose has been tilted up, but it's been losing altitude because each incremental dollar of revenue costs more, right? So what happens when the funding dries up and they're not self-sustaining entities, and they have to start letting sales people go, then the nose points down, right? So it is completely in my belief system that you will have a lot of open source companies fail, and that's a problem because one of two things happen. One is people that have deployed the software have no commercial entities supporting it, and there was never a hacker community like in the Linux world, so it either goes bust and people have software installed that there's no support for, and they kind of reap what they sow because they never paid for it. They thought it was going to go on forever, or what probably happens is the big enterprise players with distribution say, I knew this was going to happen, now I'm going to go buy the technology for scraps, and I will find a way to monetize it, and they can offer, I mean, there's a million ways to do it, right? Oh, if you want to, we have a database for content management, we have a database for a product catalog, and they don't have to sell one database per processor, they sell it much more appropriately, so the whole pricing mechanism changes, but on the pricing continuum with Oracle over here, screwing people, and the most open source, open source guys giving everything away, you can't exist on either end, the pendulum is over here, it's going to come back, so just to complete the circle, the way we're going after it at Aerospike is, we don't have a big sales force because we're not, our key advantage is not distribution. If you look at sales and marketing versus product spend at most open source companies, you know, versus ours, we're diametrically opposed, we spend way more, three quarters of our employees are either engineers or engineering focused, we have a very small sales force because we're doing a new, new thing, we're not this big deflationary pressure on the market, we actually are in a growth market, and it's harder to find and it's harder to build because it's new, but we are, we worry every day about our financials, we know what we're spending, we know what we have, we know what our runway is, and we're not going to, we're not going to hire another rep until we've made the prior rep a millionaire because he's had so much opportunity. Well, and I think investors understand, at least smart investors, understand the dynamic that you described, and they don't care because I got my investment in Cloudera, I'll get that back, or couch base or whatever it is, I just said, once an analyst, always an analyst, so I'm going to ask you sort of follow-up question on that. I was quoted yesterday in SiliconANGLE saying the space funded profitless with a lot of potential. What do you think about that? I mean, it feels like investors don't care, but the customers, I think your point is going to be left holding the bag. Well, I'm not overly sympathetic for the customers because there is no such thing as a free lunch, right? You know, it's wonderful that the venture community is finding out this opportunity, and there are going to be a couple of companies that come out of this that have unbelievable product, and it's going to be really, really good, but it's the very typical venture model where I'll fund 10, one works and pays for the other nine wipeouts, and I don't know why it would be any different this time. So are they going to be using software that's unsupported and kind of, yeah, probably, but if it's important enough, something will happen, right? I mean, if there's good technology and it's solving real problems, something good will happen. All right, Peter, we got to leave it there. Great to see you, thanks very much for stopping by. Oh, it was a pleasure. All right, keep right there, buddy. We'll be back with our next guest. Right after this is theCUBE, we're live from New York City. Right back. Live from New York. It's theCUBE covering big data NYC 2050.