>>long from Las Vegas. It's the Q covering a ws re invent 2019. Brought to you by Amazon Web service is and in along with its ecosystem partners. >>Welcome back Inside the Sands. Here's to continue our coverage here. Live on the Cube of AWS Reinvent 2019 Absolutely jam packed isles. Great educational sessions and one of the feature presenters now joins us well. Dave Alana John Walls with Paul Saville. Who's the SPP of court networking technology solutions at Caen. Freely. Paul, Good to see you again. >>Yeah, let's see you, John. >>So you just finished up. We'll get in that just a little bit. First off, just give me your impression of what's going on here and the energy and the vibe that you're getting. >>Yeah, I think it's fantastic. I mean, it's very high energy here, you know, there's a lot of new things that that are emerging terms of the applications that we're seeing the use cases for the cloud. And of course, exciting stuff happened around ej compute with the announcement of AWS with the outpost, Long >>will jump in Najaf. Everybody has a different idea, right? You weren't so I mean, if you define the edge, at least. How do you see it? >>Yeah, it's very simple definition of how we see the edge. It's putting compute very close to the point of interaction, and the interaction could be with humans or the inner action could be with devices or other electron ICS that need toe that need to be controlled or that need to communicate. But the point is getting that that computers close as possible to it from a performance standpoint that's needed. >>Okay, so we heard that a lot from Andy Jassy ethic yesterday. Right now compute to the data. I mean, with all due respect, it's like he was talking about like it was a new concept, right? We've been here for quite some time, so talk more about how you see the edge evolving. I mean, look, I have a lot of credit to Amazon because, you know, they used to not talk about hybrid. I predict a couple years to talk about multi cloud. Guarantee it because that's what customers are doing, so they respond to customers at the same time. I like their edge strategy because it's all about developers. Infrastructures code on the edge But you guys are about, you know, moving that data on or not necessarily bring in the computer that. So how do you see the edge >>evolving? Yeah, so the reason this whole trend is happening is because what's happening with the new technologies that are enabling a whole new set of applications out there? Things like What's going on with artificial intelligence and machine learning and virtual reality those the robotics control Those things are basically driving this need to place compute as close as possible to that point of interaction. The problem is that when you do that, costs go up. And that's the conundrum that we've kind of been in because when Compute gets housed at the customer premise in a home in a business in an enterprise, then that's the most expensive real estate that that there is, and you can't get the economies of scale that's there. The only other choice to date has been the public cloud, and that could be hundreds or thousands of miles away. And these new applications that require really tight control and interaction can't operate in that kind of environment, And yet it's too expensive to run those applications at the very edge at the premise itself. So that's why this middle ground now of a place and compute nearby, where conserve many locations or must be house more cost effectively. >>Okay, so you got the speed of light problem, right? So you deal with that later by making the compute proximate to the data, but it doesn't have to be like right next to it. Correct. But But what are we talking distance wise? It's that to be synchronised distance or >>when we think of the distance, we think about it in terms of milliseconds of delay, from where the edge device, the thing that needs to interact with the computer, the application needs to interact with. And we have not seen any applications that from the customers we talked to that really get beyond our need tighter than five milliseconds of delay. Now that's one way. So if we get into that range of place and compute within five milliseconds of the of the edge interaction, the device that it needs to interact with, that is enough to meet some of the most tightest requirements that we've seen around robotics control, video analytics and another >>like I could ship code to the data. But the problem is, if it needs to be real time, right, it's still too much. It's too much late, right? That's the problem that you're solving. That's right. Okay, >>so what's what you were talking about? Why milliseconds matter? That's right. So give me some examples, if you will, then about why, why five matters more than 10 or five matters more than eight or 20 or whatever, because we're talking about such an infant testable difference. But yet it does matter. In some respects. It does, >>because so give you an example of robotics, for example, robotics control. You know that is one of things that requires the most tight Leighton see requirement because it depends upon the robotics itself. If it's a machining tools that's working on a laid, then that doesn't require a tide of response time to the controller as, say, a scanning device that Israel time pushing things around very fast in doing an optical read on it to make the decision about how about where it pushes the device next, that type of interaction of control requires a much tighter, late and see performance, and that's why you get start, you start to see these ranges. But as I said, we're not seeing anything below that kind of five millisecond type of range from >>the other thing that's changing it and help me understand. This is yeah, Okay, you're moving the compute closer to the data, which increases costs. And I want to understand how you're addressing that. Maybe one of the ways addresses you're bringing the cloud model, the operating model to the data. So right patches, security patches, maintenance, things like that are reduced. Is that how you're addressing costs? >>Yeah, that is part of it. And that's why the eight of US outpost is very interesting because it is really a complete instance of AWS that is in a much smaller form factor that you can deploy very close to that point of interaction close to the customer to the customer premise, and that enables customers to leverage pretty much the full power of AWS in engaging with those devices and coding to those devices and dropping those applications closed. >>Now you lose the multi tenant aspect Is that right down unnecessarily >>from our understanding of outpost, it's a single 10 a device coming out the gate. But ultimately it's gonna be a multi tenant device. >>Yeah, okay, so near term, it's easier to manage. But it's it's multi instance, I guess, yeah, over time, maybe you could share that. That resource is still not getting. >>The interesting thing is that even though it's a single tenant device, there's still many great use cases because even a single Tenet device in set in one market could serve multiple enterprise locations. So it still has that kind of a sense of scale because you concert as long as it's it's one enterprise. Conserve many locations off of that one. That one device. >>Okay, so you don't get the massive economies of scale, but you're opening abuse cases that never existed before. >>That's right. But what about what do you do with the data supplied basically held something data scale and edge devices creating that much more data. All of a sudden speed becomes a little more challenging, taking in a lot more information, trying to process in different ways after feeding off of that, so a sudden you have a much more complex challenge because it's not static, right? This is a very dynamic environment, >>That's right. Yeah, and there's a very big trend that's happening now, which is that data is being created at the edge, and it's staying at the edge for a whole number of reasons. You know, in the Old World you would pretty much collect data and you'd ship it off to the centralized data center or to the public cloud to be housed there. And that's today. That's where 80% of data resides. But there's a big shift happening where that data now needs to reside at the deep edge because it needs to have that fast interaction with something that's that's working with or because of government regulations that are now coming in that are having much stricter tolerances around. You have to know exactly where your data is can't cross state lines. It can't, you know, get out of certain security zone. Things like that are forcing companies now to keep that massive amount of data in a very understand known localized position. >>You gotta act on it in real time. Yeah, some of it will go back to the cloud, but you see folks persist. The data at the edge or not so much persistent data. People want to store it at the edges. Well, >>uh, people in the story at the edge where where it's going to have a lot of interaction. So if you're running A if you're running a chemical plant, you may not need to have access to a lot of data outside that chemical plant. But you you're intensively analyzing that data in the chemical plant, and you don't want to ship it off someplace centrally, 1000 miles away. To be access from there. It needs to be acted on locally, and that's why it's compute this movement toward EJ computers really building and becoming stronger. >>Talk about your tech. You know what? What's the real value of what you do? You obviously reducing late, sees they gotta secure all this stuff but >>central and brings the number of tools to help in this whole space. So the first of all, the network that we provide that could tie it all together from the enterprise location to the to the edge location where compute can be housed all the way back to the public cloud core way have a network that spans the entire U. S. Fiber all over the place, and we can use those lonely and see fiber optic connections to change those those areas together in the most optimal fashion. To get the kind of performance that you need to handle these distributed computing environments, we also bring compute technology itself. We have our own variety of EJ compute, where we can build custom edge compute solutions for customers that meet their very specific SPECT requirements that could be dedicated to them. We can incorporate AWS computer technology as well, and we have way have I t service's and skilled people, thousands of employees that are focused on the space that build these solutions together. For customers that tie together, the public cloud resource is the edge. Compute resource is the network resource is the wireless connectivity capabilities that's needed on customer premise and the management solutions to tie it all together in that very mixed environment. >>We were just on a session with Teresa Carlson runs public sector for AWS, telling the SAT in a session. Marty Walsh, the mayor of Boston, has got this big smart city initiative going on. I know that's one of the cases you're working on. Maybe talk about that a little bit. And maybe some of the other interesting use cases. >>Yeah, that's right. Definitely. Smart cities are a big our big use case, though. The one and we're we're actually actively working on a number of them. I would say that those used the smart City use cases tend to move very slowly because you're talking about municipalities and long decision making cycle, I'll tell you that. We've seen >>there's a 50 year plan he put forward, >>but the use cases that we're really seeing the most traction with our interestingly is robotics is a really big one, and Video Analytics is another big one. So we're actually deploying edge used case solutions right now. In those scenarios, the Robotics one is a great one because those devices need to be. Those robotic devices need to be controlled within a really tight millisecond tolerance, and but the computer needs to be housed in a very it's much more reliable economic location. The video Analytics piece is a really interesting one that we're seeing very, very big demand for, because retailers have now reached the point with the technology where they can do things like they can, they can figure out by doing video analytics whether somebody is acting suspiciously in the store and we're hearing that they can, they think they can now cut Devery out of retail locations dramatically by using video analytics. And when you talk about big savings to the bottom line of a company that makes a big savings to them so that those very to good use cases we're seeing that a real today. You >>know what the other things you were talking about earlier was about the disappearance of Compute Divide. So where to go? Wait. >>I like to say that in the old days, if you've been around long enough like I know you're old because watching you on TV >>way get out of college, Does that make you feel way get out of college? >>Everything was in the mainframe, right? You essentially. Yet when you went to work, you had a terminal, and everything was house Essentially. Then we went to distributed where client server model, where you everybody was working on desktops and a lot of the compute was on the desk tops and very little went back to a mainframe. Then we made the ship to the cloud where he pushed his much in the centralized location as we can, too. So he's shifted way back to centralized. That's the compute divide. I'm talking about goat, that big ship from decentralized, centralized, decentralized. Now we're actually moving to a new world where that pendulum swing that compute divide is disappearing because compute isn't most economically stored. Anyone location, it's everywhere. It's gonna be at the Io ti edge. It's gonna be at the premise it's going to be in market locations. They were essential. Eyes is gonna be in the public cloud core. It's gonna be all around us. And that's what I mean by the by the disappearance of the compute >>divine. And, you know, I wantto come back on that. You talk about a pendulum. A lot of people talk about the pendulum swings mainframe and distributed. A lot of people say it's the pendulum is swinging back, but you just described it differently. It's It's a ubiquitous matrix. Now you'd is everywhere. >>That's where you hear the term fog computing the idea of the fog. Now it's not the cloud that you can see off in the distance. It's just everywhere, right, surround you and that's how combines we can start to think about how >>I first heard that you're like, I don't know eight years ago. What the heck is this? It was ahead of its time, but now it's really starting to show. This is sort of new expansion of what we know is cloud reading redefining? Yes, exactly. Net ej five g. That's, you know, another big piece of it. You know, Amazon's obviously excited about that with wavelength, right? What do you see for five G? How's that? It can affect this whole equation. >>Yeah, I think five G is gonna have a have a number of EJ applications and was primarily gonna be around the mobile space. You know, it's the the advantage of it is that it increases band with and support smoke mobility, and it allows for a little bit higher resilience because they can take the part of the spectrum and make sure that they're carving it out and dedicating it for particular applications that are there. But I tell you that the five G gets a lot of attention in terms of being how EJ computer's gonna roll out. But we're not saying that at all. edge compute is available today and that we're providing those edge compute solutions through our fiber optic networks. What we're seeing is that every enterprise that we're talking to once fiber into their into their enterprise location. Because once you have fiber there, that's gonna be the most secure, reliable and scalable solutions fiber kin can effectively scale as Bigas. Any customer could ever consume the bandwidth. And they know that once they get fiber into that application into their location that they're good for for the future because they can totally scale with that. And that's how we're deploying edge solutions today, >>Paul. I know you got a plane to catch, and you got to go. But after that age comment, we're gonna keep you for another hour. No, I think it's great. You're doing all right. All right, Hang on. We're about to say goodbye to Paul now. Well, you have a free event. 2019. Coverage continues. Right here on the right