 From London, England, it's theCUBE. Covering Discover 2016 London. Brought to you by Hewlett Packard Enterprise. Now, here's your host, Dave Vellante and Paul Gillin. Docs of Chile, London, we're here at Excel London. This is theCUBE, the worldwide leader in live tech coverage. Paul Gillin is my co-host here for the week, three days of wall-to-wall coverage. Jeff Visas here, he's the head of marketing for HP's Big Data, the vice president of marketing. Good to see you again. Awesome as always to be with you guys at these great venues. Not as intimate as the BDC. We were there, Paul and I, in August. Great show. Loved it. Yeah, that is our flagship show of the year. 2,000 Big Data people all in one room. One more could you ask for? But this show here for Europe is our biggest one and it's a chance to come as one. So, very important show. We're excited about the attendance. Spider the weather, I mean, I'm expecting it to snow any minute now, so. Coming from California, if it's not 70 degrees and sunny, something's wrong. 28 and sunny is good for us in Boston. It's hard, it's a hardship. So we were at Big Data Conference in August, speculating, of course, it was a lot of rumors swirling around about the sale of the software division that actually happened. Microfocus, we know it now. What'd you take? Give us the update on that. Well, it's exciting. It's exciting and that announcement came just a couple months ago, so it's still early days. But I can tell you, sometimes these announcements occur but because they take months, and in this case, 10 months to close, there can be this period that's limbo, where you don't get much information. But reorganizations have occurred. Priorities have been laid down and the Chris Hsu and the leadership team is right on it in preparing us for this next journey. At the same time, as evidenced by us being here, we're very much 110% part of HPE software and part of the company now. So it's a period of transition, but more importantly, a period of focus. Well, and it couldn't have helped at the time with deal flow, all these rumors floating around. Now that's sort of out in the public. How does that work? And can you bring in micro-focus to certain sales calls and say, hey, this is how it's going to operate? Look at our history, we don't mess with companies or is that something that you guys do? How has it affected deal flow in a positive way, subsequent to the rumors, so to slowing things down? Well, without, yeah, disruption your quarter end is usually not a recipe for wild success. But without divulging specific numbers as a policy we don't do it, the big data business had its fastest growing double digit growth year. So, and it didn't happen just naturally. Discussions had to give the right reassurances about the strategy, about why we were doing this and why it was good for customers and the company and the industry as a whole. Because of the focus it's going to bring with us being the sixth largest pure play software company and the fact that we are joining up with a company that owns assets like Susie, that in fact are here. A lot of people don't know that. Microsoft owns Susie, the second largest Linux provider, so not ChopLiver by any means. And that's indicative of what I observe, observe from the outside of micro-focus is when they see value, when they see success, they nurture it and let it go. Is micro-focus accompanying you on these briefings with customers to reassure them that Vertica has a good long life? Oh, well there's been, you've seen us jointly in some instances at executive levels speak about the relationship, but because the deal is not closed we operate as separate legal entities. So micro-focus certainly can and in instances as a partner. In fact they're on the show floor here. So think of them as an ISP partner today. But we do have to honor the fact that we are two separate legal entities and we have to behave that way for legal reasons and transparency reasons in the industry. We really don't compete, which makes it a lot easier. It gets a little sticky when you have two competing companies. So right now it's more about focus ahead, get prepared, and lay the groundwork. And then when the deal closes which is scheduled in the late summer that's when the companies would actually join together. So Vertica continues to do well. And as I've described, Paul, you and I have talked about this. It's sort of sat in between the sort of old cranky data warehouses and the Hadoop sort of ecosystem mess and capitalized actually on both of those. And really has solved problems. You talk to Vertica customers, they love it, it speeds things up. It's really changed many of their businesses, driving revenue. We've got dozens and dozens of interviews and examples of that. So what's new with Vertica? What's the driving this momentum? Well, we had a major release with our front loader release that came out just the beginning of September. So that is now shipping. It's being ingested by our audiences and the feedback so far is really solid. That technically quality wise, that's been a solid release. The pieces that they've gone in it, for example, two exciting ones. Of course, we added Microsoft Azure Cloud support to the support we had for AWS, for Amazon Web Services. And that is being well received that if companies want to shift some or all of their Vertica processing, they simply can take their license and run it on that instantiation. They don't have to pay a penny more. More importantly, they don't have to change the line of code. And that is a strategy we pick, which is one common core that can run anywhere. And very, very few companies have that. Can they split their workloads across private and cloud? Today you can do that in some basic ways and in the future without making any announcements that will become a lot more intelligent, a lot more dynamic. Because that's when you really take advantage of the cloud. If the cloud is all about, I pick the cloud and now I can write checks every month instead of one big check. You got some benefits to the cloud. Oh, and if I need a little more capacity, I can dial that up. Those are all good things. Those can be material. But when it gets really, really interesting is when I want to bring in new workloads and I don't want to affect my current workload, and I want to be able to have degrees of freedom of where I do that. Whether I analyze across my current Hadoop data lakes or I spin up instantiation in a cloud or go to another cloud if they have a deal of the day. That flexibility, those degrees of freedom is really what I think our long term sites are set at. When the real value of what hybrid computing brings, which it's not one or the other. And it's not lock in by the way. I mean, we talk to our customers all the time and they're saying, we've seen this movie before and we love to go to something better, but we don't love to get locked in. And everybody understands the traditional cloud tap dance which is it's free to come in, it's really expensive to leave. And I don't know if that'll ever change because there's incentives there and I don't fault anybody for doing that and if you're moving pentabytes of bits, you need a big dump track to move those down pipes or however you're going to do it. So it's not a trivial thing, but people want to preserve that ability to have that flexibility so they can have the best storage, compute and analytics and bring those together when and where they want to do that. And that's absolutely our focus. That's why our strategy to run great on the cloud, run great on premise, analyze directly when we bring it into Vertica where you get the optimized performance because we have optimized file formats. Or if you want to analyze in place, that is such a sexy term. People go, what does that mean? And it's, we can go and look at your data lake, we have optimized for parquet and or file formats. So we understand those file formats and we can analyze and hit that data without moving it. And there's a latency to doing that if the data's not sitting right where your analytic is but often for lots of use cases, that is fine. And with Vertica speed, it's more than enough. And that analyze and place capability to have Vertica be able to touch your data lakes and give you insight without changing any of your queries. That's, I think that's going to be a big home run for us. It just allows all those data lakes that have been built that people are like, I got to get some value out of them now. I mean, they're big and proud and beautiful. But if I can't get a sailboat across the damn lake, why did I build it? And some of them aren't so beautiful, right? Some of them aren't so beautiful. They're kind of messy and they need help. People didn't want to go backwards. They loved what the Hadoopika systems, what Spark and Kafka can bring. But if they still have queries to run that they need to run to run the business and they need to do a high concurrency reliably, they quite similarly don't want to take a step back. And because of where those technologies were, some were facing with that. Vertica is a really, really nice front end to that. So you can have your cake you need it to get the economics of Hadoop and get the performance of a Vertica front end. So at a big data conference, we heard a lot about Haven on demand and idle APIs. We actually played around with it a little bit. We had some use cases. We got a bunch of data. We had our developers throw some stuff at it. It was good. Some of them were a little immature. Had some ways to go. Some of them were kind of kick-ass. So give us the update on what's going on with Haven and idle. Well, Haven on demand continues to be involved and it is an open, constant beta, right? We're about two-thirds of it is commercially available APIs. If you want SLAs, it's always innovating to have new APIs there. And of course, the value prop there is simple. Runs in the cloud, any mainstream software developer can use it. So that continues to go, but we don't want to forget our crown jewel, which is idle. Idle is predetermined machine learning, enterprise class, enterprise grade. Runs on the premise can run huge amounts of data and is very, very terrible. And that's the difference with the APIs. APIs are pretty much locked in. There's a few variables, but it's kind of 80-20. And with idle, you're able to do fine grade control to help the learning and help the analysis that idle is doing around your unstructured data to nail it. And in fact, we've introduced a new version of idle and I'd love to share with you its star feature. And I even have, you know, I feel like I'm on J1 or something. So I will attempt for the camera to see the prop. And what this is, is technically it's idle 11.2. But the new feature we've added is called natural language question answering. And that's a mouthful. And this is not a new concept. The ability to talk to computers in a humanistic way, ask questions, and then have the computer reply back in a humanistic way, there have been many attempts at that over the years. Many of them failures. Clippy on Microsoft would be one. More recently there's been successes, things like Siri and Amazon Echo. And they of course use a verbal interface on top of that. But they all draw back to the same idea. Can I take a humanistic, let's say in this case, English sentence, doesn't have to be English, understand what you're asking, understand the context of it, and have a good answer. And if possible, have a conversation. Cause most people communicate like we're doing now, not with you ask me one question, I give you one answer, walk off the stage. We could do that. There might be an empty feeling for your viewers, you know, if we did that. So what we did here is, we created the ability to do something truly enterprise-grade. And what we did here is created an answer server. And how it works is very, very cool. And it harnesses idle. When you ask it a question, it's going to decide based on your question and the data that it has, one of three engines to use. The first one is called answer. Do you want to show that again? Or can we get that Leonard? Is it okay? Can we get this? Go ahead. So one of them, face this camera if you would. This, yeah, this way. Here? Okay, there you go. So the first one, I don't know if the readers or audience can read it, is answer bank. And let me explain what answer bank is. If I ask you, is the sky blue? Or I ask you, is the sky our truce? Or I say, what colors the sky and why is it that way? I'm basically asking you the same question. And you as a human being kind of know that. And I could have asked it to you 25 different ways. Well, human beings ask questions like this all the time. They don't always ask in the same way. I could be asking you what boss I should take or where should we go for dinner? You know, and we're not going to use the same words because human beings tend to communicate like that. Well, there's a concept of a reference question. And the reference question is something where we understand you're basically asking this core question. And if we can identify based on context of the reference question, we can give you a cultivated reference answer. Now, let me give you a real example. I want to program my iPhone 7 to do X. 19 different ways to answer it. Probably one right way to give the answer, okay? You know, that's pretty accurate, pretty prescriptive. Very important for Apple or any company to recognize that and be able to give that answer. The first engine does exactly that. It understands what you're really asking. It's a parsing answer. And on the information understands the prescribed answer to give to you. The second one is a little bit more direct and we call that fact bank. And fact bank allows the system to understand you're looking for a fact, okay? Something politicians probably wouldn't use a lot. Yeah, yeah, yeah. So what do I mean by fact? Maybe Facebook will. What was the price of wheat trading at on October 5th at four o'clock? It is not subject to interpretation. I need a fact. And there are places to get that information on what a stock price of trading at, what the temperature was, what the population was. And sometimes those factual information are not in databases. Sometimes they might be in an annual report and I have to pull them out. So facts can be extracted, but the whole idea is understanding what is a fact and how to deliver it when that's requested. And that's the second server. The third one is called passage extraction. That's actually very different. That's where what is the big topic of news today in London? Could you scan the top newspapers here and give me a sense of what's the big news? Because I'm going to dinner tonight. I'm not looking for a fact. There is no reference question here. I would like to extract the gist of what's going on. I may go to multiple websites. I could go to Wikipedia. I could go to BBC. But whether I'm talking about an event or an occurrence, I'm looking for the general gist of it. And I'd like you to bring that information together. These three things I've described to you are how people, 99% how they communicate. They need to do something prescriptive, something factual or something general. This system is the first that we know of that has all three engines based on high performance contextual analytics to be able to take that kind of request and produce that answer. It's not, it could be used for consumers, but it's really, really meant for the enterprise. It's meant to do serious work in the enterprise and we've released it and it's available. And I license this from you and you, yeah, what data source is? I point it to data and then it ingests it and I customize it or how does it all work? Well, exactly. Idle is, that's the beauty of Idle. As you point it at your sources, it has 13 different indexes it creates. And those indexes are basically often looking at patterns in the information. And then it will make those patterns. It can go through a learning model. So you can give it feedback to say that was a bad answer, that was a good one. And that will improve over time. And then based on the data you make available and the questions and the performance of it, this system will determine where's the best answer to give to you and then offer that up. So it's all dependent on the data you feed it. It doesn't come with an oracle of all knowledge. Now you could be hooking it up to Wikipedia and be up and running for a lot of general questions if you're getting stuff specific to your organization, how you've produced enterprise specific information you need to make that available. So are you pointing it at an HTML page and saying go ingest, are you extracting information into a database? Is it JSON format? I mean, how do you get the data into a place where it can be queried? Yes to everything you just said. That is the beauty of Idle. It has about 500 different data sources it can adjust from in a thousand different file formats. So it has an ability that we didn't invent this yesterday, this is about a seven year old technology that drives the most advanced e-discovery product offerings, the most advanced compliant archiving. So this capability is not a new one. We are riding on top of that proven capability. So it's basically expose, analyze, form the index and then start asking your questions and then tune. Can it play Jeopardy? I said can it play Jeopardy? The reason I asked that is you mentioned Siri and Echo which are consumer applications. How does it compare with Watson? Well, the users are going to have to decide on their own but just like with Haven on demand, we are often get the feedback. I can give you the feedback. I don't want to speak for another vendor. They struggle enough to speak for themselves. But on a serious note, what we get feedback is that this technology is much easier to use, much more transparent and requires far less customization. And because that's the nature of machine learning. Machine learning unlike a canned app, a word processor where you just turn it on and it starts adding numbers. Machine learning, that learning there is actually important. It is getting smarter over time to be able using these advanced pattern mapping and other cognitive functions, fuzzy logic, neural network processing. It gets smarter and smarter as it learns about what is the right answer. And that's why we call it machine learning because it is answering problems that can only be answered by learning. It can't be answered programmatically. Will there be a public proof of concept? Will anyone be able to try it? It is available for trial download right now and it is fully released as part of 11.2. So it's out there and we welcome people. People are starting to check it out already. We have some banks that are looking at it to do. The one thing that a lot of people are doing is hooking this up with some of the more popular conversation servers. That's a new genre of technology that's coming out. Our desire and strategy is to connect to them rather than compete with them. Nadia is a very famous name of one conversation server. So these are technology pieces that allow you to ask a question, get an answer and keep the conversation going and have a connected one. But the answer to your question, it's available to check out today and we've got some really great feedback about it. I think it addresses, it's the bastion that always gets it last, right? The consumers get all the good stuff because they demand it and if you don't do it better than somebody else, you're out of business. And then for some reason the enterprise where we have these highly paid knowledge workers, they're using technology from 10 years ago. And it's crazy, why are we not using natural language exchange when we're doing searching and other things in the business? It makes no sense whatsoever. Yet that's how it exists today. We're addressing it because we're putting in technologies that can do it. Because here's the thing, getting answers out in the consumer space usually it's a popularity contest. And there's nothing wrong with that. You ask a question to Google, Google has answered that question five million times and it's figured out probably what the right answer is from you because it's running this constant popularity contest. In the enterprise, you may be that first person asking that question, you may be the last. And there is not a popularity contest. And the data that you're hitting may be within your data center. It's not a harder problem to solve. Yeah, so it may sound similar and why can't I just have Siri for the enterprise and have it just start ordering groceries for me or do the equivalent. Well, there's a task that's different. And the task that's different is you're probably doing more specialized knowledge discovery and tasks. But even more of material difference is that the data you're going after is data that has to be understood in context that perhaps nobody else has already looked at before because it sits within your enterprise. Narrow, sample size. All right, we're getting the call, we're getting filed into, they're filing into the keynote to hear Meg. Jeff, thanks very much for coming back in theCUBE. It's great to see you again. Absolutely, thanks for having me. You're welcome, appreciate the update. All right, keep it right there, everybody. We're going to go to the keynotes and we'll be back after this short break, reasonably short break. This is theCUBE, we're live from London HPE Discover 2016.