 Well, welcome to this segment on Project MAGMA, which I'm going to do today with my colleague, Amar Padman-Arvin. And we've both been involved in this project. Well, Amar has been involved in it for a very long time. I've been a bit more of a newcomer to the project. And what we're going to do today is talk you through how Project MAGMA got started and why we consider it a really important piece of innovation in the wireless networking space. So Project MAGMA started off with this mission of trying to connect the next billion people. And we actually started inside Facebook connectivity where Amar used to work. And the mission statement here is to connect the world to a faster network, enabling service providers to build cost-effective, extensible and carrier-grade networks. And when you look at what it actually takes to do that, it turns out that the existing equipment and solutions provided by traditional telecom equipment vendors are not really well-suited to tackling this challenge. And the core problem that had to be addressed was one of heterogeneity. Heterogeneity raises its head in a lot of different ways. There's the challenge of dealing with different kinds of wireless. In some places you get 4G, others 5G, others unlicensed, maybe Wi-Fi, all of that kind of heterogeneity creates challenges. But there's also just the heterogeneity of all the different environments that you want to try to connect. These are not all well-resourced countries with large, highly skilled populations of people who can run telco-grade networks. And so there's this need to deal with heterogeneity of access, heterogeneity of business models, and the idea of being able to scale up, but also be able to scale down to do things at small scale in remote areas. So all of those challenges led to the thinking that a different approach was needed for building these wireless networks. And this got me thinking about innovation in the field of networking. And there's this very pivotal article written by Mark Andreessen back in 2011. Most of you probably heard of it. The article was called Software is Eating the World. And that quote has been used over and over again. It's been applied to things like how we've seen software changing the hotel industry, the taxi industry, but it's also changed things like the networking industry. And underlying that was this longer quote from Andreessen about the fact that we have all of these technologies, we have microprocessors, we have modern internet, we have clouds, we have software. All of those technologies are in place and enable us to transform all kinds of industries. But one of the things that I was perplexed about, particularly as a person having worked in networking for a long time, is why has the mobile networking industry not transformed as rapidly as other industries? And that's what we're gonna look at for the first few slides here. If you look at the mobile network architecture, you find pictures like this. And this picture could have been drawn 20 years ago because not very much was changed in the way we build mobile networks. You have radio towers, which connects to mobile handsets. You have some kind of backhaul and you have a thing called a mobile core. And the details of these changed from generation to generation as we go from 2G to 3G to 4G, but the architecture itself has been pretty static. And frankly, it has not been one that has really enabled a great deal of innovation. If you dig into the 4G architecture, for example, you see that sitting in the mobile core, there's about five different boxes which have interesting acronyms which we don't need to go into. But all of these boxes tend to be sort of refrigerator sized things that you buy from a traditional telecom equipment vendor and they are proprietary and they have a fixed set of functions determined by what the equipment vendor wants to sell you and that's what you can get. And it was dealing with this kind of lack of innovation that led the magma team to start thinking about doing things differently. Now, in parallel with what's been going on in the mobile networking industry, there's been a very rapid evolution in the way networks are built, particularly in clouds and data centers. Actually more of a revolution than an evolution. And this revolution was largely driven by this thing called software defined networking. Software defined networking, the term was coined in 2009 in this article at MIT Technology Review but it's actually an idea that's been bubbling around for a bit longer than that. And Amar and I both worked in the field of SDN when we were at a startup company called Nasera. And SDN really took off around 2011, 2012. And one of the big successes of SDN was to change the way we build data center networks. And so it used to be that you would go and buy your networking equipment in a big refrigerator sized box. It would come from a traditional equipment vendor and it would get whatever set of functions that vendor wanted to put in it. Whereas with the advent of SDN and the related technology of network virtualization, we moved a huge amount of networking functionality into software. And this now meant you could go and buy very simple networking hardware and then layer on top of it a whole lot of highly functional software that evolved very quickly and enabled all of the complexity of putting new functions into the network to be done at a much more innovative pace. And that's been really one of the big successes in data center networking. And if you look at how all the modern clouds are implemented, they've all leveraged this movement of networking functions into software. But meanwhile, back in the mobile networking world this transformation has actually not taken hold. If you look at the 4G network, it looks like this as we saw a few minutes ago. If you look at the 5G network, just got a few more boxes in it but it's the same basic architecture. These boxes do things like authenticate users, deal with mobility handoffs and you still find that they're very slow moving in terms of innovation, their functions are fixed and that's what you get. If you look at some pictures of how 5G is being rolled out, you might get the impression that it's extremely innovative. And I will say there's a lot of really exciting innovation happening in the radio side of 5G. And if you look at this picture, you'd say, oh, it looks like a modern cloud. There's H clouds and telco clouds and public clouds. It looks very, very modern. But in fact, the 5G architecture has not really leveraged all of the great technologies that have been brought to bear on modern clouds. And let me give you a very specific example. When we look at 4G and 5G networks in particular, we see this problem of abstractions and abstractions are kind of really core to how computer systems work and how we make computer systems scalable, how we make them innovative and the abstractions that have been developed for the mobile wireless network have this problem of being leaky, which is to say they don't cleanly hide the details of one piece of implementation from the rest of the system. Very specifically, if you have a 4G network, it looks like the picture on the top. If you have a 5G network, it looks like the picture on the bottom. And so I can tell by looking at the mobile core what kind of radio you're using. And this means that the radio network is leaking into the mobile core. If I can make an analogy, this would be like going to the core of the internet, looking at a great big router and trying to figure out are the users of the edge coming in over ethernet or coming in over wifi. It really doesn't matter and the core of the internet doesn't care, but the core of the mobile network absolutely cares. And so you've got this problem where the detailed choices that you make out in the radio layer leak through into the mobile core. And this turns out to have really negative implications for the complexity of the systems and for the ease with which you can innovate. And in particular, if you think about the problems faced by Magma, where what we're trying to do is we want to connect billions of people to the internet and we're going to use whatever spectrum we can get, whatever wireless technology we can get. We don't want to have to go and rebuild the mobile core every time we change radio technology. So that's one of the problems that we ran into that led to some of the design decisions of Magma. And as you'll see a bit more from Amar in a moment, one of the big innovations in Magma was to leverage the ideas of software defined networking. As I said, SDN has been one of the biggest innovations in networking in the last two decades. And it had this very important realization that centralization of control and management state generated a whole lot of benefits. In particular, once you build a logically centralized management plane and control plane, you can now think about managing a network as an entire entity, rather than as a complicated set of boxes that you manage one at a time. And rather than going and configuring a whole lot of boxes and hoping to get the desired outcome, you actually specify your desired outcome through a centralized API. And you can see that in this picture. We say, come in through this central API and say, this is the state of the network that I want to be in that creates the desired state. And then we have these quite sophisticated control planes that look at the way the world is, they collect discovered state from all of the distributed pieces that are implementing the network and reconcile the discovered state with the desired state. Another way of putting that is, we look at the way we want the world to be, we look at the way the world is and the control plane is responsible for reconciling those two to bring the state of the world into alignment with the way we want it to be. And this really changes the game for how we think about managing networks. And in particular, drives down the operational complexity of managing networks. And this is absolutely critical when we think about MAGMA as a solution that you wanna be able to roll out into under-resourced areas where you don't have a whole lot of skilled people to run the network, you need to make the operational complexity as low as possible. So those are some of the high level decisions that were made in designing the MAGMA architecture. And so now to take you into the next level of detail, I'm gonna hand over to Amar who's been involved in this in a very detailed way from the beginning. Amar, over to you. Thanks, Bruce. Over the next few slides, I'm gonna cover a couple of those principles that Bruce discussed, but a little more in detail, as well as cover like a few mechanisms that I think we leverage in MAGMA that actually make the implementation of these abstractions a reality. So the first thing that we'd like to draw everyone's attention to is MAGMA's SDN inspiration. Fundamentally, what we're doing in MAGMA is we're introducing a network level abstraction, which is the logically centralized management and control plane that is managing all of these distributed gateway devices, whether it's an access gateway, which is actually servicing the end user traffic or the federation gateway, which is interoperating with an existing core infrastructure. So the lifecycle of the actual state that is being managed in the network is fully managed through a centralized system that is running somewhere in the cloud. The second aspect of this is that, this is much bigger than what CUPS set out to be. CUPS has been trying to just separate out the control plane from the user plane, which is critical here, but it hasn't actually tried to tackle the problem of logically centralizing the control plane so as to make the enforcement of the desired state simpler. So one of the key things that we think by adding this logically centralized management from a single API and having the control plane actually enforce the state is that we think it's gonna operationally simplify the rollout of this network as Bruce covered in his section as well. The second aspect that we really try to enforce within MAGMA is we try to abstract away the heterogeneity. So as Bruce mentioned, a key problem with like trying to bring the next billion users online is a problem of heterogeneity in terms of access technologies, whether it's 4G, 5G or Wi-Fi, backhaul, whether it's fiber, unlicensed backhaul, like microwave or even satellite backhaul. The nature of the network tends to be very different. So what we try to do in MAGMA is by having these distributed edge devices, we create a interface for abstracting away the heterogeneity associated with both access as well as backhaul so the rest of the network is not affected. The other thing that we do is we leverage standard IP and internet VAN technologies like HTTP to use in inter-process communication and inter-device communication. And this allows us to build a much more robust sort of a network that can solve for both fiber backhaul as well as satellite backhaul with very long latencies. Switching gears a little. So the other big aspect that we wanna highlight is going back to the point that Bruce made around software is gonna eat the world. So a cornerstone of the principle of like software gonna eat the world is the ability to layer innovation. So one of the things that we've been doing in MAGMA is to try and leverage a lot of open source technologies and distributed system practices from the data center world as well as from like related areas like databases and big data systems to sort of solve the problems in a telecom world efficiently. A good example is currently we're using an open source project called Prometheus for our time series database. And we realized that for large MAGMA installations with more than 1,000 access gateways we're seeing some sort of performance issues in Prometheus. This is actually an issue that is observed by other projects that are not even in the telecom space. And there is a proposal to upgrade the time series database in the MAGMA project to Victoria Matrix which is yet another open source code base that has been tested at scale. So by leveraging both the learnings of other industry related projects and being able to reuse a lot of the components we're able to improve the pace of innovation in telecom infrastructure. The, a lot of the innovation that is happening today in software is predicated on two things. One is the ability to abstract away complexity. And the second one is to be able to deliver fast software upgrade life cycles. So both of these are possible today in MAGMA. For example, if we take the service here that is listed as session and policy management that is an access agnostic service that's session D in code that is, that can support Y5, 5G and 4G. So if we want to introduce differentiated policy management say, you know, like, you know traffic marking or any of these things this is something that we can introduce into the session and policy management service and make available across all of our access technologies. Further by distributing the access gateway into small components with isolated failure domains we are able to also deliver these upgrades to these devices at a much higher pace because the outage radius of like faulty software is contained to one or two like radio nodes. So reiterating this, there are two principles that add like innovation into MAGMA. The first one is our ability to abstract away complexity by building components that are access agnostic. And then the second one is to deliver these services as microservices across distributed infrastructure allowing us to upgrade these services at a very rapid pace. What are the use cases that, you know, MAGMA enables? So the current set of use cases that we support our fixed wireless access Wi-Fi access and private LTE. This half we've started focusing a lot on network expansion. This is where you wanna bring in, you know expand your existing core infrastructure into rural areas that don't support like the same cost structure that a more urban dense sort of deployment can sort of support. And we're also working in certain cases with new innovative partners like Helium to sort of enable this network expansion through decentralization as well. The bulk of our project development at this point is focused on 5G and mobile broadband. And I'll switch to the, in the next slide I'll quickly cover some of the highlights of the roadmap. Yeah, so right now we're focused on adding capabilities to the 5G essay system to make it at par with the 4G fixed wireless offering. So mostly focused on QoS and scale testing as well as integration with charging and usage reporting interfaces. In H1, we plan to work on some of the slicing foundation efforts. And this is also an area where we're looking for community help given a lot of our engineering focus at this point is on like stabilizing the 5G code base. If you would like to participate in Magma we have some list of links in the next slide. Over to Bruce to summarize. Thanks very much for that. So yeah, like I want to just sort of drive home some of the main points that Amar has made. And so the first thing here is that the challenge that Magma has really gone after is this one of heterogeneity. And as Amar said that comes in a few different flavors whether it's heterogeneity of radios, heterogeneity of backhaul or even heterogeneity of business models for the different kinds of carriers. And this is another way in which we're driving innovation because people who don't fit into the traditional model of what does a telco look like? Like helium are able to leverage Magma to start delivering innovation in terms of how they deliver wireless services. We've also seen that we can more quickly add new features into the software because of its architecture. And we're enabling innovation in things like the use of new heterogeneity devices like VR headsets that's shown here. So these two things heterogeneity and innovation really go together. And then of course the final point around scale that given a goal of connecting up the world's next billion people and beyond clearly scale is important but it's not just about making everything big. It's also about being able to scale down so that small operators dealing with rural areas, underserved areas have a solution that is cost effective for them and is something that they can operate with a relatively small group of people who are capable of delivering the services using a much more low complexity system than the ones traditionally offered by the traditional telcos. And we've got some good examples of this where Magma's being rolled out in places like rural parts of America, rural parts of Brazil. It's being used by quite a wide range of different kinds of carriers to get access to people who have traditionally been underserved by the traditional telcos. And so as Amar said, there's places you can go to learn more. I'll certainly direct you to the Magma website which has all kinds of information including details about some of the use cases, some of the companies that are using Magma, even things about like how you can get involved in the project. Also Amar and I were involved in putting together a course on edX recently so you can go there and take that free course to get a basic understanding of how the Magma architecture works and how it's different from traditional approaches to building mobile wireless networks. And then I've also wanted to mention this book that I've used a couple of times in this presentation written by one of my partners over at Systems Approach and it just gives you the background that you need to understand the core ideas of 5G architecture. And so that'll be helpful if you wanna get more involved. And ultimately that is one of the reasons that we're here presenting today is this is an open source project and it depends heavily on people getting involved and making contributions to it. And so hopefully through the little introduction we've given you today and some of these links you'll be able to figure out how you yourself can go and get involved in this project. So with that I'll say thanks very much and I believe we'll be taking Q&A through the chat. So hope you've enjoyed this presentation. Amar, thanks very much.