 From San Jose, in the heart of Silicon Valley, it's theCUBE covering Big Data SV 2016. Now your host, John Furrier and Peter Burris. Okay, welcome back, and we are here live in Silicon Valley for day two of coverage of Strata Hadoop, Big Data SV, which is our event. In conjunction with Strata Hadoop, we're right across the street at the Fairmont Hotel. This is theCUBE, SiliconANGLE's flagship program. When we go out to the events and extract the signal from the noise, I'm John Furrier with my co-host, Peter Burris, head of research at SiliconANGLE Media and Wikibon. Our next guest is Praveen Kanakari, CEO, impetus technologies and Vinit Tayagi, CTO. Welcome back to theCUBE and welcome to theCUBE first time. Thank you, John. Thank you, John. So we had a chat last time here on theCUBE about the challenges of Hadoop. And this is something that has come up. It's every show, it's like, there's gotta go faster, it's gotta be hardened. And one of the things that you guys are doing, I like to get your perspective on it is, is everyone sees the value of Hadoop. And certainly some things that change in map reduce kinda get sunsetted and now sparks in the front and center. But they understand the use case, they understand the big picture. You store a bunch of stuff in batch and then you gotta get some insights out of it. That sounds really easy, or not. So it depends who, share with us the thoughts on where you're at. And we talked about before, I wanna rehash doing it, but share with the folks out there what you guys do because you guys are in the middle of all the action. What's the update? So at a 100,000 foot level, John, one is, Hadoop came out as a very promising option for processing your unstructured data. And saw that enterprises are fully there in processing all their unstructured data. That's also a journey that they've begun on. And some are far ahead and some are behind, some have outsourced the problem to social media analytics companies and whatnot. So that's one area. Then the second area is a lot of companies started looking at Hadoop as an option for processing their structured data as well. Data warehouse or what have you. And there also there are a lot of gaps. The gaps are narrowing, conference to conference when I get to see you, but still the gaps are... It's hard, I mean it's heavy lifting involved. There's a definite skill set, there's some talent issues, but there's also some transitions we had a great chat earlier today around some of the transitional things going on around Hadoop. Certainly the ecosystem is going beyond Hadoop. The use cases are significantly sometimes vertically specialized pre-packaged apps or whatnot. The data layer, the new glue, the new middleware is there. So all these things are around where the data is stored. There's a lot of things changing in that area, how it's processed, things are changing, all flash. And then ultimately the management. Could be a nightmare for a customer to think about doing that. So given that, what have you guys found is a best practice to tackle getting Hadoop and equivalent systems out there to provide that kind of value? So largely I think what we have been doing for our customers is looking at what are these gaps? So Hadoop is a very strong contender, we believe in it, but there are gaps. And how do we bridge those gaps for them? How rapidly can we bridge the gaps? Which gaps are you seeing for instance? For example, talk of structured data. So even to use Hadoop as a viable modern data warehouse, there are a lot of things missing. And so we've actually, we've spent a lot of time and effort to examine what can we do in a hurry to accelerate this migration? Because enterprises are clamoring. I'm sure countless guests before me have said this that the cost of your legacy data warehouses is becoming unaffordable. The word we heard was retrofit. Retrofit jobs, a lot of people are retrofitting their infrastructure to kind of fit this modern era. In the other word we heard was failure. The people are, you know, need. Some are gonna fail along the way. There's gonna be a lot of gaps in what we've been seeing working with our customers as Praveen was saying earlier, is that the gaps are in several areas. One is that the distributions are there and they're doing a good job of bringing up Hadoop and making it viable. But it only comes as far. And we've been watching at the show, right? Every year the show gets bigger. 200 companies became 400 companies. And that's one of the biggest challenges that everybody has, the customers that we're working with, in order to build a viable solution and fill the gaps, they have to look at the hard part of assessing what's the right solution. And then they have to choose and work with a chosen solution they have to bring it in. And what's also happening is that for each of these companies brings a certain strength, which solves a particular gap, but leaves the other open. So you've got to then bring in some more. And then the onus is on the enterprise to manage everything and stitch everything together and make Hadoop a viable warehouse. And that's a nightmare. That's a nightmare to manage. You know, even from the perspective of these gaps getting filled up, how do you interoperate? How do you look at that? A warehouse is supposed to be a very tied together, a kind of a homogenous, well orchestrated, tied with the data, kind of a machine. And when you bring in multiple solutions, that's not happening. So you guys, your lineage is more as a services company in trying to create the actual results that people are looking for with these technologies. So effectively, it sounds like what you're saying is that you would get contracted to generate the results. Then you'd find that you'd spend an enormous amount of money with the customer in the actual infrastructure itself trying to get it to work. So you've turned that experience in the product kit that you think is just really optimized for accelerating to the results. So is that right? Absolutely, right on the money. So we were reinventing the wheel for every customer. And we would also, and we still do, we give our feedback to all the distro guys. But at some point we realized that we should be a cause in the matter, take matters in our own hands. And because they're coming from, they started at the storage layer, the computer layer, and they're coming up from there. And every year they're adding more and more abstractions and raising the bar. And we started from the customer and we started coming down. Meeting in the middle. Yeah, meeting in the middle, yeah. But our customers are benefiting from this. Now, you know, there may be some overlaps between what we do and what they're gonna do. We'll see how that plays out. So how does that translate into time to deliver? So with the old way you took how long and using your own tool set with your own people and the experience associated with it, it's happening how much faster? So going back three, four years ago, your customers were coming to us and asking us, I've got 50,000 queries to be migrated to Hive. And we would go and ask our engineers and it was a boring exercise even for the engineers working on it. I mean, 5% of the queries were challenging to migrate and required real engineering skills. The rest were just, it was a monotonous exercise. But the cost was too high. I mean, these projects were taking nine months, 10 months, and even after we had done the migration, there were a lot of challenges. The governance models exactly did not map. So it was just a nightmare. All the existing BI infrastructure could not really leverage all of this seamlessly. So fast forward, actually, we went back and over the last two years, we worked on a tool to migrate all of these workloads automatically. So now we are going into these accounts and we are doing 80% of the migration in days and weeks powered by our tool. And the 20%, which is problematic, but our tool flags the areas which require manual, smart attention and we go in and do take care of that. And so we can finish. So you automate a lot of the heavy lifting. So we just automate it. And our automation is only increasing on a weekly basis. Tell me, is that the announcement you guys had? I saw some news there. You guys have an announcement. Was that the... Yes. Vinny, can you share the announcement? So the announcement that we made, as Pramil said earlier, is that we are announcing the Data Warehouse modernization practice, which is powered by a tool set. And the tool set is proven, built on best practices, our methodology, years of our learning that we've had with Hadoop. And what it is doing for customers is that it's delivering them immediate ROI. One of the biggest area where customers are struggling is they're going on a dime and they're going on faith to invest in Hadoop. That's stuck in the mud. They get stuck in the mud. The wheels aren't turning. Exactly. So, and they're spinning their wheels. They're trying to find the right use case and business is coming and saying we invested $20 million and getting all the data in Hadoop. You've seen customers with a $20 million investment in Hadoop? Oh, yes. We've seen some of the... Overall, the program has been over a couple of years. The investment has been upwards. This is not just... This is hardware, software, and services. Total cost of ownership. All put together. And that's a tremendous amount of money which is going to be realized over a few years, right? So, this practice, this is the migration. We're talking about the tools. This is part of the practice. So, you come in and say, hey, you know, the other guys offer nine months. We do nine weeks or whatever the number is, right? Is that the kind of order of magnitude? Nine months, nine weeks? Can you... Absolutely. At least 80% on the nine weeks. Nine months or more. Is the competition? It's the competition. If you go the traditional way and just do the holding manually. And if you can hire 50 smart programmers. And you guys are in a lot of markets. He's mentioned healthcare and whatnot. So, these are verticals, right? So, you go in. You can work from the top of the stack down which you mentioned that. I like that. And there might be some gas, but you've taken the customer approach. You look at the customer. Are these pre-packaged apps? They're pre-packaged dashboards, single view into the data. Is that what they're looking for? What are these customers looking for? I mean, in these verticals? I mean, obviously, different verticals have different requirements. So, I'm sure that's where you guys add value as you come in and customize, if you will. So, in the Fortune 100, we're not going with templatized applications, so to speak. We're actually taking these tools and platforms so that we can accelerate building of their applications. Because what we're finding is their applications are, you know, their context is very unique from customer to customer. And it has to be really tuned to work in their setting. And what's been the impact that you've seen with your customers that you've gone in and done this? One of our best customers, I think we have done over 100 use cases. And what kind of impact are their business? We've got nearly 200 people working on, you know, and the impact is massive. I mean, we just did one. We moved their fraud management system, which is a massive rules-based system to a machine learning-based system running on Hadoop. And in six months, they came back and they said they had cut down $180 million with the false positives. Wow. So one is that. Instance fraud is a low-hand fruit. But then the second saving is, you know, you have far fewer dissatisfied customers who are not using their competitors. Yeah, and there's so many signals out there that they can use. So you guys are modernizing essentially. Using Hadoop as a way to get instance, they've got their attention on, I mean, who doops an indicator if the customer's saying, if Richard's Hadoop, they're looking for a new way. Absolutely. And then you say, hey, okay, we'll come in and clean up. Right, you sleep up the floor. Just so we're clear, I want to make sure I'm clear about one thing. So we have a fraud detection system today, probably running on a traditional data warehouse where the infrastructure may be five, eight, 10, 12 years old. Absolutely. And what you're doing is you're saying, hey, let's bring this application. We know it's going to become more embedded in other business activities, business processes. We know you're going to want to bring fraud or additional types of data into the fraud detection system so that you can improve its reliability, but you're not going to be able to do that on this old infrastructure. Let's move it to a new infrastructure. So the first part of the value proposition is let's move it to the new infrastructure that's going to create new options for how you embed this within the rest of your business. And then the second thing is you can take 12 months or more to do this because of the enormous complexity, or you can use this toolkit to move stuff naturally and you're getting significant paybacks. But still, how many clients can you handle at a time with this? Does it really happen that automatically or is it still relatively labor intensive on the part of your guys? It's actually today, at this very moment, it's 80% less labor intensive than it was two years ago. And we're working hard to drive it down. For you or just the customer in general. For our customer and for us, because there are not unlimited people we can hire, smart people we can hire. So even we have limitations and we are trying to drive up the value we can create per employee. And our customers are benefiting from it as well. So it's a very happy situation. So your customers get modern infrastructure, you get faster deployment and you can pursue more deals. And we can save them immediate costs, right? And they get a cheaper result. We can do it cost-efficiently, time-efficiently. Cheaper current result with new options in the future. So payback is way less than six months on a migration exercise, including the cost of, the one-time cost of migration. Got it. Yeah, so you guys, everyone wins. Yes. Well, congratulations. Great to see you again. It's very impressed with you guys last time. Again, this is a classic example of digital transformation. First step is just get the modern factory going of data. And you guys are doing that. So congratulations and I'm Praveen and Benit. Thank you so much for joining on theCUBE. We'll be right back with more. We have a big party tonight. We have a big presentation. Peter, as an analyst, we're gonna have customers. We're gonna have people in the industry and of course industry reception. Here at the Fairmont. If you're in San Jose and watching this and part of the CUBE community, come down and join us with the Fairmont Ballroom, part of Strata Headdupe. We'll be right back with more coverage after this short break.