 from San Jose. It's theCUBE, presenting Big Data Silicon Valley, brought to you by SiliconANGLE Media and its ecosystem partners. Welcome back to theCUBE. Our continuing coverage of our event, Big Data SV, continues as our first day. We are down the street from the Stata Data Conference. Come by, we're at this really cool venue, the Forger tasting room. We've got a cocktail party tonight. You're going to hear some insights there as well as tomorrow morning. I'm Lisa Martin, joined by my co-host, George Gilbert. And we welcome back to theCUBE for I think the 900 millionth time, the president and CEO of Kitting, of Data Torrent, Guy Churchwood. Hey, Guy, welcome back. Thank you, Lisa, appreciate it. So, you know, you're one of our regular VIPs. Give us the update on Data Torrent. What's new, what's going on? Yeah, I mean, we actually talked to you a couple of weeks ago. We actually did, we did a big announcement. It was around 3.10. So it was a new release that we have. Now, in all small companies, and we're a startup, you know, in the big data analytics space, there is a plethora of features that I could sort of reel through. But it actually meets something a little bit more fundamental. So in the last year, in fact, I think we chatted with you maybe six months ago, we've been looking very carefully at how customers purchase and what they want and how they execute against technologies. And it's very, very different to what I expected when I came into the company about a year ago, you know, off the EMC role that I had. And so the, although the features are there, there's a huge amount of underpinning around the experience that a customer would have around big data applications. And you know, I'm reminded of that. I think it's Gartner that quoted something like 80% of all big data applications fail. And this was one of the things that we really wanted to look at. And we have very large customers in production and we did the analysis of what are we doing well with them and why can't we do that in mass and then what are people really looking for? So that was really what the release was about. Well, so let's elaborate on this a little bit and I want to drill into something where you said many projects as we've all heard have not succeeded. You know, there's a huge amount of complexity. The sort of terminology we use is without tiring and feathering any one particular product, the open source community is really kind of like you're sort of harnessing a couple dozen animals on a zookeeper that works in triplicate. How does data torrent sort of tackle that problem? Yeah, I mean, in fact, I was desperately interested in writing a blog recently about using the word community after open source because in some respects, there isn't a huge community around the open source movement. What we find is it's the de Gère way in which we want to deliver technology. So I have a huge amount of developers that work on a thing called Apache Apex which is a component in a solution or in an architecture or in an outcome and we love what we do and we do the best we do and it's better than anybody else's thing but that's not an application, that's not an outcome and what happens is we kind of don't think about what else a customer has to put together. So then they have to go out to the zoo and pick loads of bits and pieces and then try and figure out how to stitch them all together in the best they can and that takes an inordinately long time and in general, people who love this love tinkering with technologies and their projects never kind of get to production and large enterprises are used to sitting down and saying, I need a bulletproof application it has to be industrialized. You know, I need a full SLA on the back of it. This thing has to have lights out technology and I need it quick because that was the other thing that as an aspect is this market is moving so fast and you'd look at things like digital economy or any other buzz term but it really means that if you realize you need to do something, you're probably already too late and therefore you need it speedy, you know, expedited and so the idea of being able to wait for 12 months or two years for an application also makes no sense. So the art of this is basically deliver an outcome. Don't try and change the way in which open source is currently developed because they're in components but embrace them and so what we did is we sort of looked at it and said well what do people really want to do and it's big data analytics and I want to ingest a lot of information, I want to enrich it, I want to analyze it and I want to take actions and I want to go park it and so we looked at it and said, okay so the majority of stuff we need is what we call the cash stack which is Kafka, Apache Apex, Spark and Hadoop and then put complex compute on top so you would have heard of terms like machine learning, right and dimensional compute so we have them modules so we actually created an opinionated stack because otherwise you have a thousand to choose from and people get confused with choice and I equate it to going to a restaurant into a menu on a restaurant, there's two types of restaurants, you walk into one and you can turn pages and pages and pages and pages of stuff and you think well that's great I've got loads of choice but the choice kind of confuses you and also there's only one chef at the back and he can't cook everything well so you know if he chooses the components and puts them together you're probably not going to get the best meal and then you get the restaurants you know are really good they generally give you one piece of paper and they say here's your three entrees and you know every single one of them it's not a lot of choice but at the end of the day it's going to be a really good meal. So when you go into a customer, I mean you're leaving us to ask you the question which is you're selling the prefix tasting menu and you're putting all the ingredients together what are some of those solutions and then sort of what happens to the platform underneath? Yeah so you know what you don't want to do is to take these flexible micro data services which are open source projects and hard glue them together to create an application that then has no flexibility because again one of the myths that I used to assume is applications would last us seven to 10 years but what we're finding in this space is this movement towards consumerization of enterprise applications. In other words I need an app and I need it tomorrow because I'm competitively disadvantaged but it might be wrong so I then need to adjust it really quick you know it's this idea of continual development continual adjustment but that flies in the face of all of this gluing and enterprise illities and I want to base it on open source and open source by default doesn't glue well together and so what we did is we said okay so not only do you have to create an opinionated stack and you do that because you want them all to scale into all industrialized and they don't need a huge amount of choice just pick best to breed but you need to then put a sleeve around them so they all act as though they're a single application and so we actually announced a thing called epoxy it's a bit of a riff on gluing but it's called data torrent epoxy so we have a, it's like a micro data service bus and you can then interchange the components so for instance right now a patchy apex is the stream based processing engine in that component but if there's a better unit we're quite happy to pull it out chuck it away and then put another one in now this isn't a ubiquitous snap on tool set because again the premise is use open source get the innovation from there it has to be bulletproof and enterprise illities and you have to move really fast so those are the sort of components I was working on. So Guy as CEO I'm sure you speak with a lot of customers often what are some of the buying patterns that you're seeing across industries and what are some of the major business value the data torrent can help deliver to your customers? Yeah I mean so the buying patterns when we get involved and I'm kind of breaking this down into a slightly different way because we normally get involved when a project's in flight they're one of the 80% that's failing and in general it's driven by a strategic business partner that has an agenda and what you see is proprietary application vendors will say we can solve everything for you so they put the tool in and realize doesn't have the flexibility does have enterprise illities but it can't adjust fast then you get the other type who say well we go to a distro or we'll go to a general purpose practitioner and they'll build an application for us and they'll take open source components but they'll glue it together with proprietary mush and then that doesn't then grow past and then you get the other ones which is well if I actually am not guided by anybody I'll buy a bunch of developers stick them in my company and I've got control on that but they fill around a lot so we arrive in and in general they're in this middle process of saying I'm at a competitive disadvantage I want to move forward I want to move forward fast and we're working on one of those three channels the types of outcomes and back to the expediency of this we had a telco come to us recently and it was just before the iPhone X launch and they wanted to do A-B testing on the launch on their platform and so we got them up and running within three months now subsequent from that launch they then repurposed the platform and some of the components with some augmentation and they've come out with two further applications actually three further applications you know they're all gone into production so the idea is then this fast cycles of you know micro data services being stitched together with the epoxy you know resin type approach so faster time to value lower TCO being able to get to meet their customers needs faster exactly so it's outcome based and time to value and it's time to prove because this is again the thing that Gartner picked up on is Hadoop's difficult this market's complex and people kick the tires a lot and you know I sort of joke with the customers hey if you want to obsess about components rather than the outcome then your successor will probably come see us once you're out and your groups failed and I don't mean that in an obnoxious way and it's not just you know data torrent that solves the same thing but this is the movement deal with open source get enterpriseilities get us up and running within a quarter or two and then let us have some use and agile repurposing so following on that just to understand the going in with a sort of an economic you know with a solution to an economic buyer but then having the platform be you know reusable is it opinionated and focused on sort of continuous processing applications or does it also sort of address both the continuous processing and batch processing? Yeah it's a good answer I mean in general and again geeky but you've got batch and you've got real time and stream right and so we deal with data in motion which is stream based processing a stream based processing engine can deal with batch as well so but a batch cannot deal with stream right? So you do do both and the idea being that you can have one programming model for both exactly it's just a window exactly and the other thing is a myth bust is for the last maybe eight plus years companies assume that the first thing you do in big data analytics is collect all the data create a data lake right? And so they go in there they ingest the information they put it into a data lake and then they poke the data lake posthumously but the data in the data lake is by default already old and so the latency of sticking it into a data lake and then sorting it and then basically poking it means that if anybody deals with the data that's in motion you lose because I'm analyzing as it's happening and then you would be analyzing it after at rest right? So now the kind of the architecture of choice is inject the information use high performance storage and compute and then in essence ingest, normalize, enrich, analyze and act on data in motion in memory and then when I've used it then throw it off into a data lake because then I can basically do posthumously analytics and use that for enrichment later. So you said something also interesting at where the data torrent customers the initial successful ones sort of tended to be larger organizations and those are typically the ones with skill sets to if anyone's going to be able to put pieces together it's those guys. Have you not, well we always expected big data applications or sort of adaptive applications to go mainstream when there were either packaged apps to take all the analysis and embed it or when you had sort of end to end integrated products to make it simple. Where do you think, what's going to drive this mainstream? Yeah, it depends on how mainstream you want mainstream you know it's kind of like saying how fast is a fast car? Yeah. You know if you want a contractor that comes into IT to create a dashboard, go by Tableau and that's mainstream analytics but it's not, it's mainstream dashboarding of old data. Yeah. You know if the applications that we deal with by default the more complex data they're going to be larger organizations. But it's you know don't misunderstand when I say that you know we deal with these organizations well we don't have a professional services arm we work very closely with people like HCL and we do have a jumpstart team that helps people get there but our job is to teach someone to just like a kid with a bike and the training wheels our job is to teach them how to ride the bike and kick the wheels off and step away because what we don't want to do is to put a professional services drip feed into them and just keep sucking the money out our job is to get them there. Now we've got one company who actually are going to go live I think next month and it's a kid tracker you know like a GPS one that you put on bags and with your kids and it'll be real time tracking for the school and also for the individuals and they had absolutely zero Hadoop experience when we got involved with them and so we've bought them up we've helped them with the application and we kicked the wheels off and now they're going to be sailing now I'd say in a year's time they're going to be comfortable to just ignore us completely and the first year there's going to be some still hand holding and you know covering up a bruise as they fall off the bike every so often but that's our job is IP technology all about outcomes and all about time to value. From a differentiation standpoint that ability to enable that self-service and kick off the training rules is that one of the biggest differentiators that you find data torrent has versus the tableaus and the other competitors in the market? You know I don't want to say there's no one doing what we're doing because that will sound like we're doing something odd but there's no one doing what we're doing so traditionally and it's almost like Tesla you know are they electric car or are they a platform right and you know they've spurred an industry on and Uber did the same thing and Lyft's done something and Airbnb has and what we've noticed is customers buying patterns are very specific now use open source, get their enterprise abilities and have that level of agility nobody else is really doing that the only people that will do that is your contract with someone like Hortonworks or a Cloudera and actually pay a lot of money for them to build the application for you and our job is really saying no instead of you paying them on professional services we'll give you the sleeve we'll make it a little bit more opinionated you know and we'll get you there really quickly and then we'll let you set you free and so that's one we have a thing called the application factory and so that's the snap on tool set where they can literally go to a GUI and say I'm in the financial market I want a fraud prevention application and we literally then just self assemble the stack they can pick it up and then put their input and output in and then as we move forward we'll have partners who are building bespoke applications in verticals and they will put them up on a website and so customers can come in and just download them and everything's subscription and software so Fantastic guy I wish we had more time but thanks so much for finding some time today to come by theCUBE tell us what's new and we look forward to seeing you on the show again very soon I appreciate it thank you very much we want to thank you for watching theCUBE again Lisa Martin with my co-host George Gilbert we're live at our event Big Data SV in downtown San Jose down the street from the Strata data conference stick around George and I will be back after a short break with our next guest