 Live from Silicon Valley, it's theCUBE. Covering Mobile World Congress 2017, brought to you by Intel. Okay, welcome back everyone. We're here live in Palo Alto for day two of two days of Mobile World Congress, special coverage here in Palo Alto. We're bringing all the folks in Silicon Valley, here in the studio to analyze all the news and commentary of which we've been watching heavily on the ground. In Barcelona we have reporters, we have analysts, and we have friends there. Of course, Intel is there as well as SAP and a variety of other companies we've been talking to on the phone, and all those interviews are on YouTube.com, so that's Silicon Angle. And we're here with Chuck Tato, who's the marketing director of the Data Center Communications Group at Intel, around the FPGA, which is the programmable chips, formerly with the Altair Group, now part of Intel. Welcome to theCUBE, and thanks for coming on. Thank you for having me. So actually all the rage at Mobile World Congress, Intel, big splash, and you guys have been, I mean, Intel's always been the bellwether. I was saying this earlier, Intel plays the long game. You have to in the chips games. You got to build the factories, build the fabs. Moore's Law has been the heartbeat of the industry. But now, doing more with less chips. Moore's Law making it smaller, faster, cheaper, or less expensive and just more power when the cloud does that. So you're in the cloud data center group. Take a second to talk about what you guys do within Intel and why that's important for folks to understand. Sure, I'm part of the Programmable Solutions Group. So the Programmable Solutions Group primarily focuses on field programmable Gatorade technology that was acquired through the Altair acquisition at Intel. So our focus in my particular group is around data center and comms infrastructure. So there what we're doing is we're taking the FPGAs and we're applying them to the data center as well as carrier infrastructure to accelerate things, make them faster, make them more repeatable or more deterministic in nature. And so how it works, as you were explaining before, you can set a stream of bits at it and it changes the functionality of the chip. Yes, so essentially an FPGA, think of it as a malleable set of resources. When I say that, you can create, it's basically a fabric with many resources in an array. So through the use of a bit stream, you can actually program that fabric to interconnect the different elements of the chip to create any function that you would like for the most part. So think of it as you can create a switch, you can create a classification engine, things like that. And why would someone want that functionality versus just a purpose-built chip? Perfect question. So if you look at, there's two areas. So in the data center as well as in carrier infrastructure, the workloads are changing constantly. And there's two problems. Number one, you could create infrastructure that becomes stranded. So you think you're going to have so much traffic of a certain type and you don't. So you end up buying a lot of purpose-built equipment that's just wrong for what you need going forward. So by building infrastructure that is common, so kind of cots on servers, but adding FPGAs to the mix allows you to reconfigure the networking within the cloud to allow you to address workloads that you care about at any given time. Adaptability seems to be the key thing. You know kind of trends based on certain things. So certainly the first time you see things, you got to figure it out. But this gives a lot of flexibility it sounds like. Exactly, adaptability is the key. And as well as bandwidth and determinism, right? So when you get a high bandwidth coming into the network and you want to do something very rapidly and consistently to provide a certain service level agreement, you need to have circuits that are actually very, very deterministic in nature. Shaq, I want to get your thoughts on one of the key things. I talked with Sandra Revere, I'm sorry, she was interviewed here this morning as well as Dan Rodriguez and Caroline Chan, Lynn Komp as well, a lot of different perspectives. Obviously 5G is big on one hand, had the devices out there announcing on Sunday. But what was missing, and I think Fortune was really the only one I saw pick up on this besides Sultan Engel in terms of the coverage, was there's a real end to end discussion here around not just the 5G as the connectivity piece that carriers care about, but there's the under the hood work that's changing in the data center. And the car's a data center now, right? So you have all these new things happening, IoT, people with sensors on them and devices, and then you get the cloud ready compute available, right? And we love what's happening with cloud. Infinite compute is there and makes data work much better. How does the end to end story with Intel and the group that you're in impact that? And what are some of the use cases that seem to be popping up in that area? Okay, so that's a great question. And I guess some of the examples that I could give of where we're creating end to end solutions would be in wireless infrastructure, as you just mentioned. As you move on to 5G infrastructure, the goal is to increase the bandwidth by 100x and reduce the latency by orders of magnitude. It's a very, very significant challenge. To do that is quite difficult to do it just in software. FPGA is a perfect complement to a software based solution to achieve these goals. For example, virtual switching. It's a significant load on the processors by offloading virtual switching in an FPGA. You can create the virtual switch that you need for the particular workload that you need. Workloads change depending on what type of services you're offering in a given area so you can tailor it to exactly what you need. You may or may not need high levels of security, so things like IPsec at full line rate are the kind of things that FPG has allowed you to add ad hoc. You can add them where you need them, when you need them, and change them as the services change. So it sounds, I never thought about that, but it sounds like this is a real architectural advantage because I never thought about offloading the processor and we all know we've all opened up and built our own PCs know that the heat sinks only get bigger and bigger. So and people want that horsepower for very processor intensive things. Absolutely, so we do two things. One is we do create this flexible infrastructure. The second thing is we offload the processor for things that free up cores to do more value added things. I feel like gaming for my kids loves to see that gaming. Yes, gaming, virtual reality, augmented virtual reality, all of those things are very CPU intensive, but there's also a compute intensive aspect. Okay, so I got to get your take on this. This is kind of a cool conversation because that's the virtual reality, augmented reality really are relevant. That is a key part of Mobile World Congress. So that's the IOT, which I think is the biggest story this year is IOT and all the security aspects of it around and all that good stuff. And that's really what the meat is. But the real sex appeal is the virtual reality and augmented reality. That's an example of new things that have popped out of the woodwork. So the question for you is for all these new use cases that are going to emerge, there will be new things that are going to pop out of the woodwork. Oh my God, I want to write software for that. This is an app for that now. So the new apps are going to start coming in whether it's something new and cool on a car, something new and cool on a sensor, something new and cool in the data center. How adaptive are you guys and how do you guys fit into that kind of preparing for this unknown future? Well, that's a great question too. I like to think about new services coming forward as being a unique blend of storage, compute and networking. And depending on the application and the moment in that application, you may have to change that mix in a very flexible way. So again, the FPGA provides you the ability to change all of those to match the application needs. I'm surprised as we dig into applications, how many different sets of needs there are. So each time you do that, you can envision reprogramming your FPGA. So just like a processor, it's completely reprogrammable. You're not going to reprogram it in the same instantaneous way that you do in software, but you can reprogram it on the fly whenever you would like. So I'm kind of a neophyte in this here, so I want to ask some dumb questions. Sure. I'll probably be dumb to you, but a comment to me but would be like, okay, who writes the bits? Is it the coders or is it someone on the firmware side? I'm trying to understand where the line is, between that hard and top of kind of Intel goodness that goes on algorithmically or automatically or what programmers do. So I think full stack developer or a composer or a more artisan type who's maybe writing an app. Are there both access points to the coding or is it, where's the coding come from? So there's multiple ways that this is happening. I mean, the traditional way of programming in FPGA is the same way that you would design any ASIC in the industry, right? Somebody sits down and they write RTL, they're very specialized programmers. However, going forward, there's multiple ways you can access it. For one, we're creating libraries of solutions that you can access through APIs that are built into DPDK, for example, on Xeon. So you can very easily access accelerated applications and inline applications that are being developed by ourselves as well as third parties. So there's a rich ecosystem. So you guys are writing hooks to go beyond being the ASIC special type, specials program? Absolutely. So this makes it very accessible to programmers, the acceleration that's there from a library and purpose-build. Give an example if you can. Sure, a virtual switch. So in our platform for NFV, we're building in a virtual switch solution and you can program that just like, you know, in totally in software through DPDK. One of the things coming up with NFV that's interesting enough, this is your wheelhouse right now but I want to throw it out there because it's come up in multiple interviews. And in the industry, you're seeing very cool ideas and solutions roll out. I'll give, I'll make one up off of my head, OpenStack, right? OpenStack is a great, great vision. But it's a lot of fumbling in the execution of it and it's a cost of ownership goes through the roof because there's a lot of operation on over generalizing, certain use cases, not all of the stack. But in generally speaking, Hadoop had the same problem with Big Data where great solution, but when you lay out the architect and then deploy it, there's a lot of cost of ownership overhead in terms of resources. So is this kind of an area that you guys can help simplify? Because that seems to be a sticking point for people who want to stand up some infrastructure and do DevOps and then get into this API-like framework. Yeah, so from a hardware perspective, we are actually creating a platform which includes a lot of software to tie into OpenStack. So that's all pre-integrated for you, if you will. So at least from a hardware interface perspective, I can say that that part of the equation gets neutralized. In terms of the rest of the ownership part, I'm not really qualified to answer that question. That's good. It's good media training right there. Chuck just came back from Intel Media Training, which is good, we got you fresh. Network transformation. NFD also points to some really cool, exciting areas that are going on that's really important at the network layer. You see NFD and SDN, for instance, as really important areas that people are innovating on. And they're super important because again, this is where the action is, you have virtualization, you have new capabilities, you got some security thing going down lower in the stack. What's the impact there from an Intel perspective helping this end-to-end architecture be seamless? Sure. So what we are doing right now is creating a layer on top of our FPGA-based SmartNIC solutions which ties together all of that into a single platform and it cuts across multiple Intel products. We have Xeon processors integrated with FPGAs. We have discrete FPGAs built on to cards that we are in the process of developing. So from a SmartNIC through to a fully integrated FPGA plus Xeon processor, it's one common framework, one common way of programming the FPGA so IP can move from one to the other. So there's a lot of very neat end-to-end and seamless capabilities. So final question is the customer environment. Obviously you guys have a lot of customers out there. The edge computing is a huge thing right now. We're seeing that as a big part of this, kind of the clarity coming out of Mobile World Congress. At least from the telcos standpoint, it's kind of not new in the data center area. You know, the edge now is redefined certainly with IoT and IoTP, which we're calling IoTP for people having devices. What are the customer challenges right now that you guys are addressing? And then specifically, what's the pain points and what's the current state-of-the-art relative to the customer's expectations now that they're focused on that you guys are solving? That's a great question too. We have a lot of customers now that are taking transmission equipment, for example, mobile backhaul types of equipment and they want to add mobile edge computing and NFV type capabilities to that equipment. The beauty of what we're doing is the same solution that we have for the cloud works just as well in that same piece of equipment. FPGA's come in all different sizes so you can fit within your power envelope, our processors come in all different sizes. So you can tailor your solution and then you're ready to make some. I mean, power is huge. Yes, yes. And FPGA's allow you to tailor the power equation as much as possible, right? So the question that begs the next question is, does this make it cloud-ready? Because that's a bit of a term that we've hearing a lot of, cloud-ready, because that sounds like what you're offering is the ability to kind of tie into the same stuff that the cloud has or the data center. Exactly. In fact, there's been very high-profile press around the use of FPGA's in cloud infrastructure. So we're seeing a huge uptick there. So it is getting cloud-ready. I wouldn't say it's perfectly there, but we're getting there. Well, the thing that's exciting to me, I think, is cloud, the cloud-native movement is really talks about, again, these abstractions that with microservices, and you mentioned APIs, really fits well into some of the agileness that needs to happen at the network layer to be more dynamic. I mean, just think about the provisioning of IoT. Yeah. I mean, I'm a telco, I got to provision a phone. Let's get a phone number connected to that one, then the sections go to the base station and then back to the cloud. Imagine having to provision up and down zillions of times as devices that may get provisioned once and go away in an hour. Right. That's still challenging. You need a network fabric. Yes. It is going to be challenged, but I mean, I think, as common as we can make the physical infrastructure, the better and the easier that's going to be, and as we create more. Chuck, final question. What's your take from Mobile World Commerce? What are you hearing? Analysis, commentary, any kind of input you've heard? Obviously Intel's got a big presence there. Your thoughts on what's happening at Mobile World Congress? Well, see, I'm not at Mobile World Congress. I'm here in Silicon Valley right now, but. What have you heard? Things are very exciting. I'm mostly focused on the NFV world myself, and there's been just lots and lots of. It's been a high profile. Yes. And there's been lots of activity, and we've been doing demos and really cool stuff in that area. We haven't announced much of that on the FPGA side, but I think you'll be seeing more. But you're involved. So what's the coolest thing that NFV that you're seeing? Because it seems to be crunch time for NFV right now. This is a catalyst point where, at least from my covering NFV and looking at the iterations of it, it's prime time right now for NFV. True? Yeah, it's perfect timing. And it's actually perfect timing for FPGAs. I'm not trying to give it a plug. When you look at it, trials have gone on, very significant. Lots of learnings from those trials. And what we've done is we've identified the bottlenecks, and my group has been working very hard to resolve those bottlenecks so it can scale and roll out in the next couple of years and be ready for 5G when it comes. Software-defined Chuck Dato here from Intel. Inside the queue, breaking down the coverage from Mobile World Congress. As we wind down our day in California, the folks in Spain are just going out. It should be like 12 o'clock at night there. Or going to bed depending on how beat they are. It's again, in Barcelona, Spain is where it's at. We're covering from here. And I'm also talking to folks in Barcelona. I have more commentary here at Silicon Valley and Mobile World Congress after this short break.