 Live from New York City. It's The Cube. Here is your host Dave Vellante. Hi, this is Dave Vellante and we are on the ground here at the MapR Party at the USS Intrepid. The week of Hadoop World, Big Data NYC. I'm here with Robert Novak of Cisco. Robert, thanks for coming on The Cube. Thanks for having me. It's great to be here. You, we were talking, just met. You were a big data practitioner from the early days. You talked about, you know, the six years doing Hadoop. So you know Doug Cutting. You were really one of the pioneers. So take us back. You know, what's your background in Hadoop? Well, I actually started with big data about 11 years ago before Hadoop was more than just dreams and some white papers. We were doing some parallel developments at a search company in San Francisco. I learned a lot of the concepts and didn't think about them again for a few years until I went to a company that was moving away from traditional ETL to Hadoop for the cost savings and for the reusability of the platform. So I got an early start, had a very early support contract back before. Cloudera had a ticketing system or anything. And it was a very interesting thing. I didn't know what it was going to become, but it's definitely gone farther than I expected and it's been a very interesting ride for the last five or six years. I don't know who could have predicted this, but you know, somehow you alpha geeks. You have a sixth sense as to how these things are going to take off. And I think, you know, Hadoop almost seemed like a honeypot in the early days and really attracted a lot of great brains, but it really was virtually impossible to predict how far it's come. Right. It was something that generally was a skunkworks kind of project. It wasn't something where you knew exactly how it was going to be used. And if you were lucky, you knew what you wanted it to do and you could meld it to your will, as it were. But it's only been in the last couple of years where we've gotten to the point where you can go into it knowing what to expect and actually get where you're expecting to go. And there really isn't a company out there that I can think of that doesn't have a big data play, including Cisco. So that's really your role at Cisco is to really help Cisco understand big data, learn how to exploit big data. So talk about that a little bit. What is Cisco's big data play? Well, we have a very unique server architecture that's one of our big points and one of the things that we're bringing the word out even more this week and at events like this. We have a very easy to manage, easy to scale hardware platform called Unified Compute System or UCS. And I ran on that platform for about three years with Hadoop and other analytics products and it really made a difference for me as a customer, as an end user and as the guy who got the calls at three in the morning. But another part of it is we're working really closely with our partners both in the software and the sales worlds to make people more aware of how the pieces of big data fit together, not just how they fit with our equipment, although we definitely like that. We'd like to see a lot more Cisco UCS in the world. But we want to make sure that we're connecting our customers with the companies, our partners who are going to make their businesses grow and make the most use of big data. So when you think of UCS, you think of an enterprise class, serve a platform. In the early days of Hadoop, the sort of conventional thinking was, oh, it's all white boxes. Is that changing? It is still true that you can run Hadoop on anything. You can drive almost any car from New York to San Francisco. It doesn't mean that you're going to pick every car out there. You may not want to take a smart car for that ride, even though it will technically get you there. You may want something that you know is going to get you to the end game, that's going to get you where you're going with the minimum of pain, with the minimum of inconvenience, and with the best performance you can get. One of the other things that I mentioned, you know, I have about a seven mile commute to work in the morning when I'm at home. And for me, buying a Maserati for that commute would be kind of a waste, but buying a car that I can maintain that's going to last several years is an important thing, even though I could probably buy something off of Craigslist for $300 and hope it works for a month or two. And a lot of that is the story that's around white box versus Cisco UCS, is the management story, the infrastructure, the idea that you don't have to manage, you know, you don't have to manage the systems as pets, you manage them as cattle. And you really have the ability to make better use of your resources, both your money, your people, and your time, and focus on what makes money for the business, not just what blinky lights are in the data center. So I wonder if you give us your perspective on the networking angle, because the profound thing about Hadoop when I first started following it was the notion of shipping five megabytes of code to a petabyte of data and not trying to move data. So leave the data where it is, and we're seeing, you know, a compute scale-out, we're seeing storage scale-out, does networking follow that same paradigm? And how is Cisco sort of putting forth that notion of scale-out? Where will the network go as it relates to big data? So today the network doesn't always scale out, and it's one of the big things that differentiates UCS because it's built with the network at the core of the technology. And it's one of the reasons that we can get up to 20% improvement over competing products is that we've got a high performance, relatively low latency network that just works. You know, we keep the network out of the way and just let the data flow. I think that as clusters scale, as data volumes scale, as compute requirements scale, you have to have a network that's going to support all of that traffic so that the network doesn't become the bottleneck. I've been in many environments in IT over the past 15, 20 years where the network was the bottleneck, you know, a case where going from 10 to 100 megabit Ethernet, for example, to kind of give you an idea of how long I've been doing this and how scarred I am from that, you suddenly discover that you move the bottleneck a step back and you kill your exchange server. And the idea behind a big data platform is you don't want those bottlenecks to be anywhere other than possibly in the code, and that's something that we can't quite fix yet. That's sort of in the hands of the customer themselves. But in terms of having the IO, having the memory, have the processor, and having the network that scales seamlessly, whether you're doing a 10-node cluster or a 10,000-node cluster, that's a really big thing. That's something that everybody in the market is going to have to deal with from a hardware standpoint. And I think we at Cisco have a real advantage there because, you know, it's what we're known for is our networking technology. Robert, I'd love to continue this conversation on another Cube activity, another Cube interview. We've got to leave it right there. Thanks so much for coming on. This is Dave Vellante, and you're watching The Cube. Thank you.