 Well good morning all right what a beautiful place huh couldn't believe it when I was driving here yesterday wow what have you've never been here before I want to talk to you about something that's really exciting that's going to really impact all of us whether you're a developer part of a small company part of a big company whatever you are in the next few years 5g is going to impact you personally and your business and I want to talk about so what does it mean this this era of 5g and where does open source come into this let me start by this is probably a slide you've seen it before just to give you an idea of of how this the story of going to software and open source started for us at AT&T about four or five years ago when when we deployed the iPhone in 2007 in the first ten years of the deployment of the iPhone our traffic increased by 360,000 percent majority of our traffic to actually more than 50% of our traffic today is video the traffic certainly has changed quite dramatically and every day we carry this is data as of February 2019 we carry 242 petabytes of data daily huge amount of data and we saw that coming few years ago and what we did about this is to start really disaggregating our network into a control layer and and a data layer we started sort of moving sort of the software and the hardware and trying to understand how to use software not just to scale our network more efficiently but to program our network so we started building this platform that called Ecom and three four years later we worked with Jim and Arpit and decide you know what we want to partner with the Linux Foundation partner with a number of companies China mobile and many other vendors to really put that into open source and today it's we have a very thriving on app ecosystem of over a hundred companies under the Linux Foundation networking umbrella this has really helped us significantly today at AT&T we've deployed on app this is an open source we've deployed at AT&T and we've actually by the end of 2018 we passed 65% of our network being virtualized our target network AT&T 65% virtualized today managed by this orchestration platform this is great so what's next we're starting to see big shifts we're starting to see a shift also not in just now traffic but in the devices we're going to more smartphone devices and we're expected to have an explosion of IoT devices or explosion you know due to the 5G an explosion of the M2M and M2X what this means is that that little device that you probably have in your house it could be a toothbrush or whatever will probably have the intelligence of your smartphone today that couple of dollars maybe a dollar even we're talking about well if that toothbrush has the smart the smartness of your smartphone today can it collect data and send that data to your dentist if they're you know to tell them whether there's something wrong with your teeth or they have cavities etc we're not far away from that we're actually two three years away from that with 5G basically coming in it's all going to be basically possible so there's an explosion shift in the types of devices and the number of devices that are going to be that we will have to connect in the network and the current approach that we have today is really not going to scale to allow us to support that the type of data is also changing we started many years ago with just text and that's gone to voice to images to video and with 5G we're expecting a huge part of the traffic is going to be virtual reality augmented reality video 360 etc starting from text this data explosion is about 50,000 times and that 50,000 times is not something you can grow a network by 50,000 times just in a span of about 10 15 years so we're starting to look at that so well how is the network going to support this change this explosion of data and the shift in the devices and the data most of you have seen the evolution of the network okay 1g you can support you know 2 to 4 kilohertz kilo bits per second you know we use that to its analog we can basically transmit voice with that then we went to 2g and with 2g now you've gone into not just voice you can actually do some rudimentary data now you can have 10 to probably 14 kilo bits per second and now you can start some doing some some rudimentary texting and voice and then you go to you go to 3g and in 3g we started seeing sort of a movement from voice and data to doing more data than we've ever done before we can probably support something like two three megabits per second of data but really the big revolution really happened with 4g and that data became video and that really gave birth to a lot of companies that even in existence today we would never be able to do streaming video as one example if it wasn't a 4g and LTE and I think what you're seeing in the 5g this is really a paradigm shift and this is why 5g is so important because when you really put together 5g and what we heard earlier today about edge compute you put that smartness at the edge and you increase the speed of the access really what you end up with a real-time network what does that mean a real-time network means that you can now do the compute the intelligence the processing in the network just as fast as you can have it in the device you have an enormous large amount of intelligence sitting just at your fingertips wherever you are and that's why this is really a big deal 5g sort of allow us to operate at spectrum and we've talked about millimeter wave spectrum what does that mean is that you're now operating at spectrum about 24 gigahertz per second gigahertz or above that kind of spectrum that kind of frequencies really allow us to do sort of this tremendous bandwidth in that type of frequencies so you see us moving towards 5g a millimeter wave very quickly as we deploy 5g at a scale globally this also means that latency is going to be reduced we're talking about 1 to 2 millisecond latency for 5g and now when you marry that with edge compute and really trying to put GPUs and other high-performance compute at the edge then you're looking at round speed round trip type of compute and processing of something like maybe 10 to 20 millisecond now compare that with a hundred to 200 milliseconds that you'll basically get today and that really trying to drive a paradigm shifts in the type of application you can run now some of these applications are pretty obvious autonomous vehicles autonomous vehicles can never be scalable and mainstream without really the advent of 5g and edge compute without that real-time network to be able to collapse and process data across multiple cars and do it right in real-time to avoid crash to you know manage the traffic you need that real-time network all the way to autonomous drones we started the trial autonomous drones without a captain and with video capability that can help us we were talking about AI earlier today Jim was having a session on AI but we're applying AI in the network today in the network and one place we're applying it we are feeding drones with visual capabilities to go up and tell us whether our cells have a problem or not whether it's dirt one of the number one problem with all cell sites is dirt okay basically to tell us whether we need to maintain the cell side or not now remember we have 75 just us we have about 75 80,000 macro cells and that's gonna be hundreds of thousands of small cells as you go to 5g 5g is a paradigm shifts from these macro big cells to very small cells that you can pretty much position every mile or two so an explosion the type of applications so what does this mean to what we've done over the past few years as we've brought to this to this level is that we've gone to software we are now a big advocate of software we use a lot of software actually in AT&T we counted in the past two years we have contributed 10 million lines of software just in two years which is quite astonishing for a company certainly like AT&T which is a very traditional telecom communication company these are the softwares we're starting to use today in building 5g and today I'm not gonna go into a lot of the details as to how and why but I'll just give you a little glimpse of what's going on if you're not familiar with some of those software you're probably familiar with the open stack this is what we use for cloud compute you know for a lot of our cloud compute technologies and this is what we have deployed in our cloud zones at AT&T today airship is what we use for the management of these cloud services and it's based on Kubernetes and then you get Daynoss is something we announced last year and this is sort of think of it as your network operating system for white boxes and this is something very vital to us because as we start thinking of the edge the edge is not just some cloud zone sitting few miles away the edge could also be where your customer is where your factory is where your retail store is so we're starting to deploy a lot of these white boxes with a network operating system called Daynoss here. O-RAN you're gonna hear a lot about O-RAN so O-RAN is a well-defined foundation it's led by a number of operators and a number of vendors are part of that and the goal of O-RAN is to really to disaggregate the RAN and drive interoperability you'll hear more about that in the next month or so. O-RAN is one of those capabilities we're using in building the edge for us as AT&T we found that with one of our most challenging tasks AT&T when we are building these multiple edge zones is integrating the technologies so O-RAN is about how do you basically drive and and the integration of edge technologies in a way that we could sort of test and deploy as a cookie cutter. Acumus is part of the Linux Foundation deep learning umbrella we're using Acumus in building analytic and predictive capability for the edge in figuring out how we can load balance our traffic so we load balance our traffic video traffic and others we have to have the predictive capability to allow us to know which cell which carrier do you basically load balance your traffic and this is basically what we use we use AI capability and Acumus to do exactly that and ONAP is our network operating system for the network cloud. All of these capabilities is what we're using today as we are rolling out 5G. I thought I'll show you the slide and you're gonna see more about the slides moving forward one of the most if you want to work on a complex task the big complex task to work on is the RAM this is really probably I've worked on a lot of software and open source I've never worked on something as complex as the RAM and part of that because it's really a close box it requires a lot of expertise in data and radio engineering and and many other expertise that you know it's hard to find one person who has all of these expertise together and what we've been doing as part of the ORAN consortium is to really break down the RAM into smaller pieces. You'll see basically at the bottom is your radio unit this is the little device that you see on top of a tower and today that little device is completely connected to a one big box called the VBU and what we're doing here is that opening up the box into smaller software modules with open API interoperability and standardizing those APIs and you can see where we're using the open source across all of these across that basically stack what's phenomenal in here is latency latency is very very important when we've been working in own app you know operating at about doing a close loop in the order of a second to two seconds or even higher it's okay in the network now if you want to identify there's a failure in your virtual machine or you want to drive traffic from one place to another apply a different policy it is okay if you're operating at a half second or a second or more but when you are looking at access really you're now talking about 10 to 20 milliseconds for things doing close loop within the the CU and and really in the sub millisecond when you're going all the way down to the radio units so now you know the expertise that you have to have is more than software it's more than data is you really need to understand how to manage latency and speed so this is just another view of looking at this is we talk about you know AT&T about this concept of edge-to-edge and this is where we are basically all that means is that we are placing the intelligence there's no one physical location for edge so we talk about edge compute we don't talk about one place we really talk about many places that start with the device all the way to the customer premise where we are deploying these white boxes and could be in your home and could be in small enterprises and then you have the cell towers of basically putting some of that intelligence in your cell towers like I said we're going from macrocells to microcells this is the movement towards using these millimeter wave spectrum and now also placing some of that intelligence in these edge clouds being in the data center so this is your CEOs your national data centers your regional data centers and what we're doing with those is that we're starting to push the intelligence from not just the device itself or what is today it's just the device itself and the core cloud the third-party cloud we're starting to redistribute that intelligence all across to get the optimal speed latency and experience basically that you want to support the SLA that you want to support for the application and then you have your private cloud and third party cloud and you could see where the open source software that we are placing across the across the stack from edge to edge one thing that we don't talk more a lot about and Martin mentioned this in the previous talk I know this is a software conference we talk a lot about open source okay but really what's also just as exciting as software and it's so important is the data and the reason why data is important because for us to optimize and use software in an intelligent way it's really through data and that's why AI is interesting because it's all about really bringing data and software together to try to do something really magical so we talk about 5g you hear a lot about 5g but one thing I want you to get out of my talk is when people talk about 5g I won't talk about latency and speed and new experiences and virtual reality and augmented reality and autonomous cars this is all good and this is all great but I want you what to remember is that the data and the question is that we're gonna be we're gonna have an explosion of data with 5g and one thing that's really exciting for a community like us here is what we gonna do with this data what kind of new projects we can have in Linux Foundation by marrying the data with software I want to show you I want to show you a video here that tells you a little about how we are starting to think about using AI and data in our development of 5g oops can you play at AT&T we're no stranger to artificial intelligence and machine learning we've embraced these technologies as valuable tools in our ever-evolving toolbox and are leading the way in integrating them into our world our next use 5g first AI will help us build out our 5g infrastructure it maps out where cell towers fiber lines and other transmitters exist today and can pinpoint the best location for 5g buildouts pairing location and usage information will create the most optimal 5g network machine learning and AI will work together to keep the 5g network maintained and secure by continuously monitoring how it's being used and how it's responding so if one cell is not functioning properly AI will signal another tower to pick up the slack or if one area is experiencing a high volume of usage AI will trigger lower use cell sites to ensure speed isn't compromised so what does this mean for you an explosion of data a strong efficient 5g network built for a mobile workforce a world of autonomous vehicles near zero lag virtual and augmented reality experiences from across the globe and truly extraordinary possibilities so we're actually using open source software today in designing our 5g so this is an actual use case is that every small cell AT&T puts is not based on an army of engineers going out and figuring out what to do and talking to the town etc it's really based on machine learning and AI technology that really ingest a large amount of data and tells us basically where is the most likely place we can put a small cell and what the experience is what the throughput is and how this is going to impact our customers another use case that you saw in the video is that now you put these small cells in how do you make sure you load balance those cells the traffic across those cells because for those who understand 5g and small cells one of the challenges with small cells and 5g is the propagation is that you have very short propagation and you have issues about penetrating through walls and glasses and things like that so again optimizing that again the same type of problem is that we're applying some of the open source technologies that I mentioned earlier and really building more data driven type of predictors to tell us how do you position antenna and what angle you can position the antenna to get the best experience okay so I mentioned this I'll just finish with those two use cases we're really very very excited about this new paradigm shift with 5g and it's really not about 5g it's about bringing 5g and edge together into this old you know real time network it's about bringing AI and pushing AI not in the traditional AI but really pushing AI closer and closer to the customer so at their fingertips they have really that intelligence that compute that power to do a lot of things with it wherever basically they are it's really about making the network programmable and that's what software defined network is if the network is not programmable it's really very hard you can have all the speed that you want it's really very hard to change things you know in in a millisecond type of interval to change spectrum in a millisecond interval to change the bandwidth that you would need in a millisecond type of interval so we're applying all of these we have actually opened up several foundries in the United States and one in Israel and one in Mexico and we've we've we've opened up sort of 5g and an edge that uses some of these open source software in these foundries to allow us to prototype with small businesses and large businesses new concept in 5g so if you want to try something you want to test something you you have an idea and open source or some new software come and work with us in our foundries whether it's for the autonomous cars and we're working with a number of car manufacturers where we are really pushing that intelligence to the small cell and to manage a number of cars sort of moving at the same time or even redefining the retail of being able to monitor manufacturing and really change the conveyor belts and manufacturing based on the point of sale at that moment in time using 5g technologies so let me finish here and thanks very much for listening and look forward to working with everyone here