>>from around the globe. It's the Cube with digital coverage of AWS reinvent 2020 sponsored by Intel and AWS. Yeah, husband. Welcome back to the cubes. Live coverage of AWS reinvent 2020. We're not in person this year. We're virtual This is the Cube Virtual. I'm John for your host of the Cube. Roger Barker, the General Manager AWS Robotics and Autonomous Service. And a lot of other cool stuff was on last year. Always. Speed Racer. You got the machines. Now you have real time Robotics hitting, hitting seen Andy Jassy laid out a huge vision and and data points and announcements around Industrial this I o t it's kind of coming together. Roger, great to see you. And thanks for coming on. I want to dig in and get your perspective. Thanks for joining the Cube. >>Good to be here with you again today. >>Alright, so give us your take on the announcements yesterday and how that relates to the work that you're doing on the robotic side at a w s. And where where does this go from? You know, fun to real world to societal impact. Take us through. What? You how you see that vision? >>Yeah, sure. So we continue to see the story of how processing is moving to the edge and cloud services, or augmenting that processing at the edge with unique and new services. And he talked about five new industrial machine learning services yesterday, which are very relevant to exactly what we're trying to do with AWS robot maker. Um, a couple of them monitor on, which is for equipment monitoring for anomalies. And it's a whole solution, from an edge device to a gateway to a service. But we also heard about look out for equipment, which is if a customer already has their own censors. It's a service that can actually back up that that sensor on their on the device to actually get identify anomalies or potential failures. And we saw look out for video, which allows customers to actually use their camera and and build a service to detect anomalies and potential failures. When A. W s robot maker, we have Ross Cloud Service extensions, which allow developers to connect their robot to these services and so increasingly, that combination of being able to put sensors and processing at the edge, connecting it back with the cloud where you could do intelligent processing and understand what's going on out in the environment. So those were exciting announcements. And that story is going to continue to unfold with new services. New sensors we can put on our robots to again intelligently process the data and control these robots and industrial settings. >>You know, this brings up a great point. And, you know, I wasn't kidding. Was saying fun to real world. I mean, this is what's happening. Um, the use cases air different. You look at you mentioned, um, you know, monitor on lookout. But those depend Panorama appliance. You had computer vision, machine learning. I mean, these are all new, cool, relevant use cases, but they're not like static. It's not like you're going to see them. Just one thing is like the edge has very diverse and sometimes mostly purpose built for the edge piece. So it's not like you could build a product. Okay, fits everywhere. Talk about that dynamic and why the robotics piece has to be agile. And what do you guys doing to make that workable? Because, you know, you want purpose built. The purpose built implies supply chain years. in advance. It implies slow and you know, how do you get the trust? How do you get the security? Take us through that, please. >>So to your point, um, no single service is going to solve all problems, which is why AWS has has released a number of just primitives. Just think about Kinesis video or Aiken. Stream my raw video from an edge device and build my own machine learning model in the cloud with sage maker that will process that. Or I could use recognition. So we give customers these basic building blocks. But we also think about working customer backward. What is the finished solution that we could give a customer that just works out of the box? And the new services we heard about we heard about yesterday were exactly in that latter category. Their purpose built. They're ready to be used or trained for developers to use and and with very little customization that necessary. Um, but the point is, is that is that these customers that are working these environments, the business questions change all the time, and so they need actually re program a robot on the fly, for example, with a new mission to address the new business need that just arose is a dynamic, which we've been very tuned into since we first started with a device robo maker. We have a feature for a fleet management, which allows a developer to choose any robot that's out in their fleet and take the software stack a new software stack tested in simulation and then redeploy it to that robot so it changes its mission. And this is a This is a dialogue we've been seeing coming up over the last year, where roboticists are starting to educate their company that a robot is a device that could be dynamically program. At any point in time, they contest their application and simulation while the robots out in the field verify it's gonna work correctly and simulation and then change the mission for that robot. Dynamically. One of my customers they're working with Woods Hole Institute is sending autonomous underwater robots out into the ocean to monitor wind farms, and they realized the mission may change may change based on what they find out. If the wind farm with the equipment with their autonomous robot, the robot itself may encounter an issue and that ability because they do have connective ity to change the mission dynamically. First Testament, of course, in simulation is completely changing the game for how they think about robots no longer a static program at once, and have to bring it back in the shop to re program it. It's now just this dynamic entity that could test and modify it any time. >>You know, I'm old enough to know how hard that really is to pull off. And this highlights really kind of how exciting this is, E. I mean, just think about the idea of hardware being dynamically updated with software in real time and or near real time with new stacks. I mean, just that's just unheard of, you know, because purpose built has always been kind of you. Lock it in, you deploy it. You send the tech out there this kind of break fixed kind of mindset. Let's changes everything, whether it's space or underwater. You've been seeing everything. It's software defined, software operated model, so I have to ask you First of all, that's super awesome. Anyway, what's this like for the new generation? Because Andy talked on stage and in in my one On one way I had with him. He talked about, um, and referring to land in some of these new things. There's a new generation of developer. So you gotta look at these young kids coming out of school to them. They don't understand what how hard this is. They just look at it as lingua frank with software defined stuff. So can you share some of the cutting edge things that are coming out of these new new the new talent or the new developers? Uh, I'm sure the creativity is off the charts. Can you share some cool, um, use cases? Share your perspective? >>Absolutely. I think there's a couple of interesting cases to look at. One is, you know, roboticists historically have thought about all the processing on the robot. And if you say cloud and cloud service, they just couldn't fathom that reality that all the processing has cannot has to be, you know, could be moved off of the robot. Now you're seeing developers who are looking at the cloud services that we're launching and our cloud service extensions, which give you a secure connection to the cloud from your robot. They're starting to realize they can actually move some of that processing off the robot that could lower the bomb or the building materials, the cost of the robot. And they can have this dynamic programming surface in the cloud that they can program and change the behavior of the robot. So that's a dialogue we've seen coming over the last couple years, that rethinking of where the software should live. What makes sense to run on the robot? And what should we push out to the cloud? Let alone the fact that if you're aggregating information from hundreds of robots, you can actually build machine learning models that actually identify mistakes a single robot might make across the fleet and actually use that insight to actually retrain the models. Push new applications down, pushing machine learning models down. That is a completely different mindset. It's almost like introducing distributed computing to roboticists that you actually think this fabric of robots and another, more recent trend we're seeing that were listening very closely to customers is the ability to use simulation and machine learning, specifically reinforcement. Learning for a robot actually try different tasks up because simulations have gotten so realistic with the physics engines and the rendering quality that is almost nearly realistic for a camera. The physics are actually real world physics, so that you can put a simulation of your robot into a three D simulated world and allow it to bumble around and make mistakes while trying to perform the task that you frankly don't know how to write the code for it so complex and through reinforcement, learning, giving rewards signals if it does something right or punishment or negative rewards signals. If it does something wrong, the machine learning algorithm will learn to perform navigation and manipulation tasks, which again the programmer simply didn't have to write a line of code for other than creating the right simulation in the right set of trials >>so that it's like reversing the debugging protocol. It's like, Hey, do the simulations. The code writes itself. Debug it on the front end. It rights itself rather than writing code, compiling it, debugging it, working through the use cases. I mean, it's pretty different. >>It is. It's really a new persona. When we started out, not only are you taking that roboticist persona and again introduced him to the cloud services and distributed computing what you're seeing machine learning scientists with robotics experience is actually rising. Is a new developer persona that we have to pay attention to him. We're talking to right now about what they what they need from our service. >>Well, Roger, I get I'm getting tight on time here. I want one final question before we break. How does someone get involved with Amazon? And I'll see you know, whether it's robotics and new areas like space, which is verging, there's a lot of action, a lot of interest. Um, how does someone engaged with Amazon to get involved, Whether I'm a student or whether I'm a professional, I want a code. What's what's the absolutely, >>absolutely, so certainly reinvent. We have several sessions that reinvent on AWS robo maker. Our team is there, presenting and talking about our road map and how people can get engaged. There is, of course, the remarks conference, which will be happening next year, hopefully to get engaged. Our team is active in the Ross Open Source Community and Ross Industrial, which is happening in Europe later in December but also happens in the Americas, where were present giving demos and getting hands on tutorials. We're also very active in the academic research in education arena. In fact, we just released open source curriculum that any developer could get access to on Get Hub for Robotics and Ross, as well as how to use robo maker that's freely available. Eso There's a number of touch points and, of course, I'd be welcome to a field. Any request for people to learn more or just engage with our team? >>Arthur Parker, general manager. It is robotics and also the Autonomous Systems Group at AWS Amazon Web services. Great stuff, and this is really awesome insight. Also, you know it za candy For the developers, it's the new generation of people who are going to get put their teeth into some new science and some new problems to solve. With software again, distributed computing meets robotics and hardware, and it's an opportunity to change the world literally. >>It is an exciting space. It's still Day one and robotics, and we look forward to seeing the car customers do with our service. >>Great stuff, of course. The Cube loves this country. Love robots. We love autonomous. We love space programming all this stuff, totally cutting edge cloud computing, changing the game at many levels with the digital transformation just a cube. Thanks for watching

>>From Las Vegas, it's the cube covering AWS reinvent 2018 brought to you by Amazon web services entails their ecosystem partners. >>Okay. Welcome back everybody to the cube live in Las Vegas for AWS Amazon web services reinvent 2018 upshot four with David, Dave, our sixth year covering AWS reinvent. We EV except for the first year we weren't there, but certainly it's been fun to watch the massive, massive rive of the wave of the cloud and Amazon's discipline and execution. Our next guest is Roger Barga, general manager, robotics and autonomous services for Amazon web services. Great to have you thank you for joining us. It's great to be here today. So a lot of stuff to talk about this, Amazon's got like this cult personality, or they do cool things. Uh, they innovate as well as they take care of the basic cloud needs more compute, better networking, more storage, or the core engine, uh, robotics, autonomous, you think of cars, you think of future flying drones, maybe in the future. >>What's going on? What are you, what are you working on? I think it mentioned explain what your job is and what you're doing at Amazon. I think it's super important. We actually look at robots as being anything that census computes and acts, and that opens up such a wide range of the definition of robot from a washing machine to escape the system to the robots. We think of actually that's the full spectrum is what we're trying to address. And we've announced a new service called AWS robo maker. It is designed to support the end to end application development life cycle for building intelligent robot, deploying it to one 10 hundreds, thousands of robots out in the field, monitoring them. We are really addressing the developer need on how to build and scale and run a robotics business. You know, what really resonates with me and, uh, with you guys at Andy's keynote this morning was he used the word builder a lot of times, um, tool for the right job. >>I think that really connects with the culture that we're seeing in the world today. Maker fair started it out. Robotics clubs in high schools were probably at an all time high in terms of interests. It's not just a nerdy geek thing. It's actually kind of mainstream. People are attracted to rabbis. People have wearables. So you're seeing a world where technology and robotics are colliding. So this kind of falls into the new kind of persona developers that's out there. Who's building a robotic stuff. It used to be some like special group of people. Not anymore. Explain how you guys are going after the developers with this. Okay. So it is very focused on the developer. And we started talking to our internal customers who are building robots. We started talking to external customers, building robots to really understand the struggles that they had and have to face. >>And you actually realized that the roboticists tend to actually are deepened hardware, drivers, actuators, sensors, and they are forced to be software engineers at the same time, because there's just not ready-made software and they have to go roll their own tooling. So we're actually providing them with the tools so they can actually focus on the hardware and the innovation that goes on there, or adding the intelligence to the robot to carry out the more meaningful task. And again, we've had conversations with companies that are, that are building small appliances that basically they think of as a robot, a dishwasher that has sensors, they've actually sense how the water flow is going the temperature and then take action all the way to our group. That's actually putting a robot in the space station to take photographs all over underwater robots, air robots, and the drones. So those deed came in robotic competitions, right? >>You're familiar with those, right? It was all high school kids. And there's always a hardware team, which is kind of clear. And then the software team, which always struggled. So I'm envisioning these guys are now going to be using robo maker as part of that team. So if I understand it, the mission is kind of develop secure, deploy, and manage robotic apps. That's really what you guys are a little bit more also, please. So we've actually bundled in our cloud service for machine learning, for analytics and for monitoring. And so now with Amazon Polly and Amazon Lex integration, you can talk to your robot, your robot can respond to you. We can stream the video off the robot through Kinesis, video streams and send it to recognition. So the robot can actually see, you'll be able to see what your robot is seeing, run it through recognition. >>You can identify what it's, what it's seeing and be able to tell it, go to the refrigerator. And it knows where the refrigerator is something else we have done. I think it's interesting to share with you is that we've actually working with something called the robot operating system, which is the most commonly used open source software framework for robotics ROS. Um, we have contributed all of our cloud extensions as open source to the community. And we're also technical steering committee members for Ross two, which is the next generation of Ross. We like to think of it as a commercial grade version of Ross, the Linux for robots. And we're also contributing open source to that as well, because what you'll find is this is what developers are using and reusing. So if you have a sensor or an actuator for a robot you'd like to use, you're probably going to find ross' package already out there to actually drive that sensor or drive that actuator that you can use. >>And now you see new ones for our cloud services that you can turn monitoring on machine learning services on as well. So you contribute to open source community you're so that's going to accelerate the adoption. So you're also making it easier. I want you to explain how you guys are working to do that because if this kind of continues on this track is going to remove some of the blockers or the barriers to get into this and that's to get the applications up and running, which should have a impact on like fleet management to, you know, anything. I mean, that's really the problem statement here. Isn't it, it really isn't, it's really what our mission is. We're always looking at developers and how we can accelerate them and make them more productive. Let's say the three of us wanted to go off and build a robotics application. >>We'd have to make sure that the environment and all of our machines are the same, because you might have a DLL, a different DLL or a different package, which means when we deploy to the robot, we're breaking it. We're not consistent. We actually offer a cloud development environment for robotics. With one click off the AWS management console. You can choose the operating system that you'd like to deploy to your robot. It'll download it. It'll configure that for you. It'll create scalable storage to store the artifacts. As we build our robot and try different algorithms out it'll provision compute for, to compile our, our robot application. We even have pre-built applications to get you started and you have access to all the ROS packages. And so within minutes we could it be up and working together, writing a robotics application. That's just part of it though. >>So again, I talked about the cloud service extensions, but simulation is such a huge thing because we may not even have a robot bill yet. And we want to simulate our robot. We offer pre-built worlds like a room in a house or a retail store or a racetrack for the race car that you heard about today. And you can drop your robot in these environments and test it. You can turn a physics model on and say, my robots carrying 500 pounds simulate. When you're happy with it, then you can deploy that over the air to your actual robot and the simulation. You can actually run hundreds of them in parallel, faster than wall clock time. So it's literally, we could actually do a thousand simulation hours, probably in 15 or 20 minutes to test our robot and all this compute, you spin up a supercomputer, basically bring it all together. >>You mentioned the formula. One thing, that's interesting. What insights can come into this. And I want to get down to the intelligence piece because when I met Andy, I just wrote an article yesterday on Forbes with my, on my interview with him, he made a comment. I want to add to the conversation. He said, the clouds are the brains on premise as their environment. So robots will deep rains. So talk about the connection to the AWS. Yes. So that's a key part, right? It connects to the, they got a lot of brains. So you got a lot of opportunities to connect services. What kinds of services do you envision connecting to the robots? Okay. So what was announced today with the race car it's at that car is actually trained in robo maker through simulation, through reinforcement learning. And so hundreds of simulations of the car, trying to go around the track, all that information is being fed to SageMaker, which is using its reinforcement learning to actually build an algorithm, a better algorithm, and then pulling it back to the car and trying it over and over again. >>That's how you actually train the car and you see that beautiful partitioning with the cloud, big compute, reinforcement learning, large datasets. The car wants you to deploy the machine learning model to the car. It can actually continue to set up signals for more information. So as the car is being used for racing, you're still learning. It's still updating the model. So again, this beautiful part, how's that how's that data flow. So you have data coming off the car, you send it back up to the cloud, you then that's where the heavy modeling occurs. And then you push it back down. The small machine learning model, back down, we have Kinesis data streams. We also have IOT MQTT messages. We can send back up to the cloud and you really start to see the role of the cloud. When we have hundreds of devices out, each one might make a mistake every once in a while, but collectively you're getting a large training set for returning a model and pushing it back down. >>It's where deep learning really adds value, too. It really is. And you mentioned adding more personality to it before we came on camera robot, you saw, this is really kind of where it's going to really kind of make it personalized. It, it is. And in fact, Leah, it's this it's a robot that's made by by robot care systems, excuse me, robot care services. And Leah is an intelligent robotic Walker. Absolutely brilliant. The elderly and disabled canal live more independent, more agile lives. Um, it has 72 sensors since compute act. It figures out what the user is trying to do. The user now can actually interact with it with voice through our Amazon Polly and Amazon's Lex integrations. So with the walkers across the room, the user can say, Leah, come to me and Leah will actually motor over to the user user can get on. >>Leah will sense that it's carrying load and it can say, Leah, let's go to the front door and Leah will start moving our way to the front door. That's just so natural. And that's the impact of real life impact of that. People who live alone, could it be diabetes or maybe something as they get sick robot could be tied into a health meter. I mean, this is kind of real world scenarios that aren't far away. No they're happening now. It's happening right now. And again, you're starting to see the value that robots are going to bring to our lives. And again, robotics has to have such hard problems to solve with the hardware and that algorithm, the writing. We really don't want the other work to have to be a burden for them. We really want to simplify that. So I'll talk about the CHAM, the total market adjustability here, because the F the formula one, the developers, I get that Jennifer's I get the formula one. Is there a market for robots? Who's doing it. Where is it? Is, is it embryonic and early? Is it, how's this forming you in your mind? Um, marketplace, as we've looked at this, we have been amazed at all the places we're finding robots. Again, we see robots underwater. We see drones in the air. We see robotic arms and factories. We see them in education. I have yet to see an area where a robot can assist or carry out tasks to help humans. How about doing interviews? >>Yeah. We're not gonna be replaced yet. Although we have >>Robot on the cube one, despite the fact that we'd like to think how advanced robots are, you can't replace humans, not the NR, the mobility, our intelligence or personality. So if the number of things robots could do keeps getting, >>Yeah, it wasn't, it wasn't that long ago, robots couldn't climb stairs. >>That's right. That's right. Amazing. Let's talk about your goals for the year. What are you trying to do with the, with the service? Um, and what can people expect to see coming from AWS? We're definitely going to be listening to our customers now that we've launched and we're working backwards to actually add features that they tell us. They'd like to see. We're really pleased that we've got a partnership with first robotics. We want to work with with first, actually bring our service to allow students and learners of all ages to learn robotics. We have an education and research program with about 25 universities with more signing on as well. They're very interested in using the service for teaching robotics and for education and research as well. So I really want to, we really want to push hard there's because we think robotics has a great future. >>It's going to help our lives. And we think robo makers, the way that they're going to do, I can tell you from my four living in Palo Alto, which is again, a different zip code than middle America, robotics is hot. People like robotics. They like to play with the robotics. And it has now it's software democratization tools and frameworks. You don't need to be a rocket scientist to code sheet language. Yeah. Yeah. That's I think the power of our service is that basically the developers no longer limited to the code. They write in the software. They can hardware that can put on their robot that can take advantage of cloud services, glue them together and start building a robot. Well, we are very interested in covering, uh, what goes on with your area and certainly want to know more about how the community's developing. Certainly the open source I think, is going to be a very big part of your plan. We agree. We're committed. Roger. Thanks for coming on. Great insight, robo maker. One of the top announcements is a great demo on the keynote, uh, from, uh, the formula one, uh, spokesperson. I think the executive great demo that I think is worth watching. Congratulations on the success or cube coverage here. No robots here. We're live coverage. Re-invent 2018. We right back.

>> Narrator: Live from Washington D.C., it's theCUBE covering .conf2017 Brought to you by Splunk. (techno music) >> Well, welcome back to Washington D.C. We're at the Walter Washington Convention Center as we wrap up our coverage here of .conf2017. As Dave Vellante joins me, I'm John Walls here at theCUBE, coming to you live from our nation's capital. Joined by Team AWS here. With us we have rather, Ray Zhu rather, who is a senior product manager at AWS. And Roger Barga, who is the general manager of Amazon Kinesis Services. So gentlemen, thanks for being with us, we appreciate the time. >> Absolutely, thank you for the invitation. >> Dave: Oh, you're welcome. >> You bet. Alright, so let's just jump in. The streaming data thing, right? It's just blowing up. What's inspiring that popularity of the Cloud? What's kind of lit that fire and what's going to keep it burning? >> Yeah, I think over time, I think customers really do realize the value that you can get out of by collecting, analyzing, and reacting to data in real time. Cause that really provides a very differentiated experience to their customers, you know, for example you're able to analyze your user behavior data in real time, provide them with a much more engaging experience, much more relevant content. You're able to diagnosis your service, understand your law of data issues in real time, so that when you have an issue, you can fix that right away. So that really provides a very different customer experience. So I think our customers are realizing the value of real time processing, which is why we think streaming data is gaining more and more popularity. And this is why Cloud is all the good stuff that Cloud can offer and tell the customers. It's highly scalable, so you don't need to worry about if it's going to scale later on when I scale my business. It's a matter of sort of like click of a button. We scale the infrastructure for you and we got all the resource ready for you to go on streaming data. We got super, it's very cost effective, right? So that cause we price at very low. As we keep improving the efficiency of running the service, we reduce our cost structure, we return that back to our customers as a price cut. The third thing which I think is super important is agility, right cause you don't need to set up an infrastructure, install any software, make all the configurations. Starting up a Kinesis Stream is like 15 seconds on the average console, you're done. And it really allows the developers, the customers, to move fast and purely focus their resources and effort on the things that really differentiate their customer experience. >> So very AWS like, we love AWS, we're a customer, it's our favorite Cloud. We'll go on record of saying that, you know? (laughs) We're loyal to you guys. Crowd, our Crowd Chat App runs on it, basically run our whole company on Amazon, where we can. >> Roger: Great. >> In 2013, we got the preview of Kinesis. It was a lot of buzz. It was kind of before the whole streaming meme took over. We were talkin' about real time at the time, but so maybe you can take us through the evolution of Kinesis and where we are today. >> I'd be happy to. You know, when we first built Kinesis Stream, what the company was trying to do, is we had all of the AWS billing and metering records coming from all of our services, our EC2 incidences. This was a lot of data that had to be captured. And the way we were doing it was in batch. We were storing this data in S3 buckets. We were starting large EMR jobs up at the end of day actually to aggregate them by the customer account. So say this was your bill for the end of the day. But we had customers that said actually I'd like to know what I'm spending every hour, every few minutes. And frankly that batch processing wasn't scaling. So we had to innovate and create Kinesis Streams as a real time system that was constantly aggregating all of the billing and metering records that were coming in from our customer's accounts. Totalling them in near real time and we presented our customers with a new experience of billing and insights into their billing and even forecasts of what they were spending at any given time. But we had other teams that immediately looked at Kinesis and said hey, we're dealing with real time streaming data and our customers want it delivered and aggregated and provided, so Cloud watch logs and Cloud watch metrics built on top of us. And this was the start of something which continues to this day. Other services are looking at, and even customers, are looking at a Kinesis Stream and saying, that's a really useful abstraction that we can build a new service, a new experience for our customers. And today we have over a dozen AWS and Amazon retail services that build on top of Kinesis Streams as a fundamental abstraction to offer new experiences and new insights as three events. Cloud watch events, there's a host of services, which underneath Kinesis is running, but they're offering unique value building on top of it. Which is why Kinesis today is considered a foundational service and we can't build an AWS region without Kinesis being there for all these other services to build on top of. So that's been exciting to see that kind of adoption, different uses for this fundamental abstraction called a Kinesis Stream. And you know, it's also, and we can talk later about how it's transforming analytics, which is really exciting as well. >> Well, that's a great topic. I mean, why don't we talk about that. And one of the things that we've noted about AWS, and other Cloud providers, is obviously simplicity and delivering as a service is critical. We all know about the complexity of, for instance, the Hadoop Ecosystem And the challenges that a lot of customers have. Delivering that as a service has dramatically simplified their lives. That's why you see so many people going to the Cloud. We've always predicted that is what happened. Maybe talk about that a little bit. And then we can get into the analytics discussion. >> Yeah, so again, customers are always looking at ways to actually get insights into their data to better support their customers, to better understand what's going on in their business. And of course, Hadoop had managed EMR, had been a great benefit, cause customers could move their developers into the analytics that they want to do and not worry about this undifferentiated heavy lifting of operating these services. And the same is true for Kinesis Streams. But we're seeing customers, and if you stop for a moment and think about this, data never loses it's value. It always has it's historical value for machine learning, for understanding trends over time, but the insights that data has are actually very, very perishable and they can actually turn to zero within an hour if you can't extract those insights. That's the unique area where Kinesis Streams has kept adding value to our customers. Giving 'em the ability to get instant insights into what's going on in their business, their customers, their business processes, so they can take action and improve a customer experience, or capitalize on an opportunity. So what we're seeing and the role, I believe, that streaming data, at large, plays is about giving customers real time insights and then business opportunity to improve how they run their business. >> So. >> Go ahead, please. So who's using it? I mean or what's the if there's a sweet spot or a sweet spot for an industry or vertical to use that, I mean, in terms of whether it's in a minute, an hour, or whatever, what would that be? >> Yeah, so today, I'm really pleased to see, because we have watched this evolution since 2014, but today in virtually every market segment, where data is being continuously generated, we have customers that are actually taking advantage of the real time insights that they can get out of that data virtually every market segment. I'll pick a couple of examples which are kind of fun. One is Amazon Game Studios, near and dear to our heart. Now typically games are written, they're completely developed end to end. They're shipped in a box, made available to customers, and they hope that game and the engagement has the outcome that they want. Amazon Games Studios is actually writing that game in near real time ahead of their customers, so they release a new level of the game. They will actually watch the engagement. They'll look at how customers are dying, surviving, how long they're playing. And is it traveling in the direction they want? They stream all of the multi, all of the game data from their players in real time. And they build dashboards so they can see exactly how game play is going. And if they don't like it or they think they can make an improvement, they'll get right online, change the game itself, and re-deploy the game, so the customer experience is actually, within minutes it's being evolved. Another customer I like to talk about is Hertz Publishing. We all like to read. When Hertz started making the transition of their magazines, Cosmopolitan, Car and Driver, from print to digital form, they instrumented it so they could actually watch how long was a customer reading an article, how were their comments trending in Twitter and in Facebook. So they could actually get a sense of engagement with an article. Whether the article should be rebroadcast to other digital channels, other magazines. Should they change the article? Double down and write a new one. So again, they're engagement and then the business metrics by which they measure engagement and readers, readership have all increased because they have that intimate understanding of what's happening in real time. So again, every market segment, where there's data continuously generated, customers are using this to provide a better experience. >> That phrase undifferentiated heavy lifting we first heard it widely in the tech community in 2012 in Andy Jassy's keynote at Reinvent and it's become sort of a mantra. It probably was one well before that inside of AWS. And often times AWS doesn't talk about TCL but it's not the main reason why people go to the Cloud. You emphasized that a lot. And there's all this debate. Oh a cheaper on prem, oh no, Cloud is cheaper. But this idea of essentially eliminating labor that is doing that non-differentiated heavy lifting is something that you guys have really lived and popularized. We see that labor cost shifting from provisioning luns into other areas, up the stack, if you will. Application, digital business, analytics, et cetera. What are you guys seeing, in terms of how organizations, I mean, there's two types of organizations, right, the Cloud native guys who obviously didn't have the resources, but then enterprises that are bringing their business to the Cloud. Where are they shifting that undifferentiated heavy lifting labor towards? >> To. And they are in fact moving it up stream. We think about it very abstractly. You know, operating servers doesn't really bring any special IP that that company possesses to bear. It is about, you know, just managing servers, managing the software on it, figuring our how to scale. These are problems which we are able to take away. And we've often worked with customers and showed them the value of moving to our managed servers. And the excitement from the leadership, from their customers, is like wonderful. That project we couldn't, we aren't able to fund, if we can just onboard here, onto Kinesis for example, or any one of our managed services, then we can immediately move and get that fund project that we really wanted to fund, it would actually be unique value as move them over to that. So they're actually moving upstream as you said. And they're actually leveraging their unique understanding of their industry, their customer, to go ahead and add value there. So it is a distribution and I think in a very productive way. >> I want to ask about the data pipeline. So one of the values that AWS brings is simplification. When I look, however, at the data pipeline, it's very rich. If I look at the number of data services, Kinesis, Aurora, DYNAMO dv, EBS, S3, Glacier, each of these has a programming interface that is, I use the word primitive not in pejorative way but >> Roger: Yes, yes. >> But a deep level, low level. And so the data pipeline gets increasingly complex. There's probably a benefit of that, because I get access to the primitives, but it increases complexity. First of all, is that a fair assertion on my part? And how are your customers dealing with that? >> Be happy to take that one, yeah? >> Sure. >> Okay. >> Yep, so I think from our perspective all these different capabilities and technologies by customer choice. We build these services because our customers ask for them. And we order a wide variety so that people can choose for the developers who want to have full control over the entire staff, they have access to these lower level services. You know as you mentioned a few, DYNAMO dv, Kinesis Stream, S3, but we also build an abstraction layer on top of these different services. We also have a different set of customers asking for simplicity, just doing a specific type of things. I want you guys to take care of all the complexities, I just want that functionality. The example would be services like Kinesis Files, Kinesis Analytics, which is the abstraction layer we put on top. So for customers who are looking for simplicity, we also have these kind of capability for them. So I think at the end of the day, it's customer choice and demand. That's why we have this rich functionality and capabilities at AWS. >> So you guys have already solved that problem essentially, the one that I was sort of putting forth. >> So I won't say, I like Ray's answer. It's about listening to the customer. Cause in many cases if we would have, if we said, hey, we're going to go build a monolithic service that simplifies this, we would potentially disappoint many other customers. Say actually I really do want to have that low level control. >> Right. >> I'm used to having that. But when we hear customers asking for something which we can then translate to a service, we'll build a new service. And we will actually up level it and actually build a simpler abstraction for a targeted audience. So for us it's all about listening to the customers, build what they want, and if it means that we're going to actually bring two or three of our services together to work in concert for our customer, we'd do that in a heartbeat. >> Yeah that low level control also allows you to be presumably maybe not more agile but more responsive to the market demand. Because if you did build that monolithic service, you would essentially be locking yourselves in to a fossilized set of functions and services that you can't easily respond to market conditions. Is that a fair way to think about it? >> That is a fair statement, because basically our customers can look at these API's and together for these various services, realize how to use these API's in concert to get an end and done. And should we have precise feedback on a specific service, we can add a new API or tailor it over time. So it does give us a great deal of agility in working on these individual services. >> So Ray, you're a product guy and you're talking about listening to customers, right? And coming up with products, it's what you do. What are you hearing now? Where do people want to go now? Because I assume you've been in the market place for four years now with this, evolution is (clears throat), excuse me, perpetual, constant, so where do you want to take it? What's the next level or what's percolating in the back of your mind right now? >> Yeah, I think people always looking for different type of tools that they're familiar with or they want to use to analyze these data in real time and provide a differentiated customer experience. A concrete example I want to give is actually why we're here. At the Splunk Conference is at Kinesis we have a service called Kinesis Firehose. Based on customer demand when we launched Kinesis Streams, customers wanted to make sure they had access to data sooner than they used to do, but they want to use the tools they're familiar with. And apparently there's a diverse set of tools different customers want to use. We started with S3 for data lay, kind of storage, we used Reshift as a data warehouse. And overtime we heard from customers say, hey, we want you to use Splunk analyze the data. But we would like to use Kinesis Firehose and suggest a solution. Can you guys do something about it? So actually the two teams got together. We thought it's a strong customer value proposition, great capability for other customers. So we start this partnership. We're here actually earlier this day, today, we made the announcement actually, Kinesis Firehose is going to support Splunk as data of redestinations. And this integration is not in beta program. It's open for public sign up. Just go to the Kinesis Files website. You can sign up, get early access. So basically from today, you can use Kinesis Firehose in real time streaming (mumbles) service to get real data into your Splunk cluster. We're super excited about it. >> And okay, and I can access those Splunk services through the market place or what's the way in which I bring Splunk to? >> Good question. For this integration actually we're just a different version of Splunk. You can run Splunk on AWS using ECT extensions. You can access through the market place. You can have your, you can use native Splunk Cloud, which manage all the servers for you. You can also use Splunk on print in that regard. >> Okay. What have you guys learned since the orig, the first reinvent? I mean, I think, and again, I don't mean this as a pejorative but AWS is pretty dogmatic in its view of the world as you you are very strict (laughs) about your philosophy. But at the same time, as you learn about the enterprise, you've evolved. What have you learned about enterprise customers in that five, seven year journey of really getting intense with the enterprise? >> Yeah, that's a good question. But again, we're dogmatic about we always listen to our customers. We will never deviate from that. It's part of our culture. And the customers need to tell us where they want to go. And I'll tell you when we first started with Kinesis, just to answer your question, it was about low latency. We want to get that answer really fast, cause our ad tech customers are some of our very early customers, so it really was about that that extremely low latency response. As even our customers have started to look at Kinesis as a fundamental abstraction on which to put all of their business data in and now they're telling their customers well you should, if their IT customers within their company, if you want any business data, attach to the stream and pull it out. So now we're seeing less emphasis on low latency and to end processing, but increase request I want to be able to attach a dozen consumers, because this stream is actually supporting my entire enterprise. I want to have security. So we recently released encryption at rest. Our customers are asking for support for a VPC flow logs, which we hope to be talking with you about very soon. So now it's becoming actually very mainstream to actually, for the enterprise, and they want all the enterprise ready features, all the certifications, Fed Rep, Hippa, et cetera. So now we're actually seeing the Kinesis Stream itself being put into the enterprise as a fundamental building block for how they're going to run their business and how they're going to build their applications within the business. >> So that philosophy, I mean, you are customer driven first and there's a lot a, Andy Jassy says, there's a lot of ways to compete. You can be competitive oriented, but we're customer oriented. And I, it's clear, you guys do that. At the same time, customers sometimes don't know what they want, so you have to be good at decoding. >> Roger: Yes. >> If you listen to all your customers, you know, five years ago, they say, well we're not going to put any data in there. Sensitive data in the Cloud. Now everybody has sort of gotten over that. You said, alright, well we have to make it more secure. We have to get, you know, whatever certified, et cetera, et cetera. There's an art to this, listening to customers, isn't there? >> It gets back to one of our leadership principles of we always work customer backwards. We need to understand what they want, what experience they'd like to have. We have to anchor everything on that. But there is this element of invent and simplify. Because our customers may guess at what a solution is, but let's make sure we really understand what they want, what they need, the constraints under which that solution must offer. Then we go back to our engineering teams and other teams and we invent and simplify on their behalf. And we're not done there. We actually then bring these back to customers and in fact, why we're here today, we've spent two days talking to customers but even before this collaboration with Splunk began, we actually brought customers in and it turned out, their customers were often our customers. So we started talking, what is the problem? And we started with the very clear problem stain. And once both of our teams, we've loved working with Splunk, they work very customer backwards, like we do. And together once we understood this is the problem we are trying to address, and we had no preconception about how we're going to do it, but we worked backwards on what it would take to actually get that experience for our customers. And we're actually here beta testing it. And we're going to have a very aggressive two or three month beta test with customers, did we get it right? And we'll refine as well before we actually release it to the customer. So again, that working with the customer, work customer backwards. But invent and simplify on their behalf. Because many Splunk customers weren't aware of Firehose until we explained it to them as a potential solution. They're like ah, that will do it, thank you. >> So very outcome driven. I mean, I know you guys write press releases before you sometimes launch products. Sort of as you say, that's what you mean by working backwards, right? >> Roger: Yes, yes it is. It really is. >> Ray: You're good listeners. >> So far it's worked. (laughter) >> It's always fun at the company, when somebody says I have a customer, the entire room gets quiet and we all start listening. It's actually fun to see that, because that's the magic word. I have a customer and we all want to listen. What do they want? What are they challenged with? Cause that's where the innovation starts from which is exciting to be part of that. >> It's been a great formula, no doubt about that. >> It has, it has. >> Thank you both for being here. Didn't realize it was a big day. So congratulations >> Thank you. >> on your announcement as well. >> Absolutely. >> Ray, Roger, good to see you. >> It's great talking with you. >> Alright, you're watching theCUBE live here from Washington D.C. .conf2017. (techno music)