 I welcome this opportunity to share some of my thoughts with you. Some preliminaries, I'm a wireless guy, so from that perspective I will speak. And secondly, we're gazing into a crystal ball, so images out of a crystal ball are not guaranteed to be accurate. So, with that in mind, let's begin our journey. I think I like to start a little bit with an analog to Moore's Law for the network, which is Cooper's Law. Cooper is a guy who came out of Motorola, and basically Cooper's Law says every 30 months or so, the capacity of the network doubles. So I plotted that on the left for you, going back to the early days of wireless in a logarithmic curve that looks like a straight line, so there's some exponential growth there. But I don't want to focus so much on the shape of the curve, but I want to talk a little bit about what caused all this capacity increase. If you look at it, almost all of it, and when almost all, I mean 80 to 90%, some authors have wrote as much as 95% is due to densification. What does that mean? That means in the early days of wireless, we take a transmitter, we blast it out into the world as far as it will reach, and then after a long, long, long distance, we use that frequency again. And then somebody in Motorola got a great big, good idea in around the mid-70s and said, why don't I limit the power and therein created cellular? So basically now you have, take this vast transmission distance and densify the number of transmitters, making it smaller. And then we've been on that journey ever since, making the cells smaller and smaller and smaller. So basically almost all the gain in capacity is from densification. Now, of course, there's gains from spectrum increase. Spectrum increase, as you know, will gain linearly through Shannon bound, but not exponentially. Of course, there's things that are moving closer to the Shannon bound and we've been doing that. And I'll talk more about the Shannon bound in a little bit. Now let's shift over to the other side of the figure here. The right-hand side, I'm left-right dyslexic, so I can't tell left from right. So over here, the last speaker touched on it already, that we have exponential growth in the data that's been carried through the mobile network. And that's shown in the red curve there. But what's also interesting is if you look at the mobile revenue, the mobile revenue is almost flat. So I started showing this curve in 2010 and got into a very depressed funky mood because what that means is that I got to start building infrastructure that is exponentially cheaper in a cost per bit. That's not a good thing. So, but eventually, I come to the conclusion, if you are only about reducing costs, that's the minute your industry is dead. So we cannot just be reducing costs. I need to bend that top line up. So today, we're going to talk about how to bend that top line up. And I'll give you some ideas in my head about how to do that. Incidentally, within this, what's happening in those two curves, we have the development of SDN and NFV. I don't want to define and talk a lot about it. There are talks today at this conference about NFV and SDN. The only thing I want to say is NFV is the running of network function, non-generic hardware, and SDN is the separation of control and data plane. And on top of that, we have this open source community, which you guys belong to. So our great hope is that all of you will be able to help us bend that curve up. And looking back a little bit, the iPhone was a tremendous back swan into our industry. Tremendous change. But it has to come from Apple. It could not come from one of us. You know, Steve Jobs loved the fear of aluminum. So he wanted to build his phone with an antenna inside his aluminum case. You come to us, our F guys, we say, oh, that's stupid. You basically put your antenna inside a Faraday cage. Good luck with your RF transmission. And, you know, and we have seen that in the early days of the iPhone. How you hold the iPhone impacted the performance of the iPhone tremendously. So that's because you got this antenna inside a Faraday cage. But eventually they solved it. What does this tell us? It tells us that you need to come at things from a fresh new idea. Those of us who are like me old and dragged down with dogma, that's not kind of where we need to be. So how do we bend that curve up? Well, I think in order to bend that curve up, what we need to do is create a new network that will allow us to create virtual network to open applications. Now, over the top players have been sending packets through the wireless network for a long time now. But one thing I will tell you is the wireless network compared to the wire network, I like to say it's actually a broken network. And why is it broken? In a sense, if you think about it, the wire network, you know, my guys in Optical tells me they can achieve 10 to the minus 12 bit error rate. If I'm in the wireless world, I'll be ecstatic if I can achieve 10 to the minus 3 bit error rate. That's the difference of roughly 10 to the minus 9 or 10 to the 9. To give you a sense of what that is, it's like the thickness of a human hair compared to the diameter of the earth. That's how different those two networks are. So in a sense, you know, the network in the wire line world is perfect and the one in the wireless world is imperfect. It's imperfect because there is no wire so that we have to go through the air and the air interface matters in that transmission. But given that the air interface matters, the only entity that knows about the air interface is the wireless network. But that entity is not giving that information to the over top applications. How does over the top application work today? It works with trial and error. It sends a bunch of packet out and hopefully one of them will get to you. Now, having said all that, that will give you a sense of how inefficient it is today. But if I could open up my wireless network and provide you using data analytics, certain aspect or information about my network and how data are going through my network, then I hope all you brilliant people in the open source community can make use of this information and build significantly better applications that are much more efficient. I need efficiency because I cannot live with this exponential curve. I need to somehow dampen that down too. But by doing this, you create a whole new revenue that I need to bend the curve up. More importantly, you're going to think of whole new business cases that I have even thought of. I only thought of one with giving you specific virtual network that are personalized. All right. I did all this and I haven't even talked about the air interface yet. I told you the air interface is very important. So let's have one little graph on it. Spectrum is scarce. It's been scarce all its life. So that isn't going to change anytime soon. So we need more spectrum. And where do we find big chunks of spectrum? Big chunks of spectrum is up in the millimeter waves. So we're going to need, in the future, to take advantage of all available spectrum. That's one thing. Second thing is we all know about Wi-Fi. We have Wi-Fi here. So Wi-Fi or unlicensed spectrum is always going to be with us. But we should not make it do what is not intended to do. I'll tell you a story. VTC4 happened in September of this year. The keynote speaker is the CTO coming out of Telus, the biggest tier one operator in Canada. Unfortunately, the night before, something drastic happened in the Telus network and he has to stay home, work on that problem. Maybe that's better. He has to stay home, work on that problem, and couldn't travel to Vancouver to give the keynote. So there we are, the day of the keynote. So we try to run a teleconference from where he was in Calgary to Vancouver and try to run it through Wi-Fi. And those of you who try to run Wi-Fi with a room that this many people will know, it ain't going to work very well. And indeed, it didn't work very well. The video conference was terrible. And so this comes to one of my proudest moments in my career. Basically, what happened is that they switched to LTE running through my network, my infrastructure into one of my devices, and that worked very well. Now, the lesson here is not one is better than the other. The lesson is take advantage of all the spectrum you have, make it do what is intended to do. So there'll always be, unlike in spectrum, we'll need to make use of it. Third thing, dynamic spectrum sharing is coming. This comes from the idea that, you know, if you look at government spectrum, a lot of it is not really used. Think about it from the Navy point of view. They have a bunch of spectrum for their radar. So, you know, they use that word. There is ocean. But in the middle of the U.S., Kansas, there ain't no Navy there. There is no Navy there. So that spectrum isn't used at all. So the idea here is to dynamically share the spectrum so a secondary user can actually use it when the primary is not using it. And this, on the right-hand side, has just released in July several studies that look at the impact of LTE onto radar. And it shows that the impact of radar to LTE is not very much. And that's to be expected. Radar is a very narrow beam. It spins around. And it comes around every once every five seconds. So losing a packet or two every five seconds, no big deal. On the other hand, if you look at the red curve, if I have any kind of significant power in my LTE to the radar, it causes significant impact to the radar performance. So any kind of LTE transmission is terrible. So I'll tell you, dynamic spectrum sharing, it's a big thing. But how we share, that needs to be worked out. That needs to be worked out by a number of very bright people. It's going to be the future. But when we'll get here, it's a good question. I talked to FCC guys and they tell me it'll take them five to 10 years to get the regulation down. So it's going to be a little bit. But it is going to change the whole landscape. All right. What should the future network support? On this box, I plot what we call the hypercube, the surface hypercube. And in the middle of that hypercube there, you see what the cube looked like today. So basically what it says is that in the future, you're going to have applications that is going to be much broader, require much bigger range of throughputs, much bigger range of delays, and much bigger range of links. All right. If you want more about how this hypercube came about, go find our white paper on the web. Now, the other thing I want to say is signaling traffic when compared to data traffic. We saw a lot about data traffic growing exponentially. So I plot data traffic and signaling traffic together on the same curve. Admittedly, they're at different scales. So don't take those numbers too hard. But what you can see is that data traffic is growing much, much faster than, sorry, signaling traffic is growing much, much faster than data traffic. And that hasn't taken into account by what's going to happen when we expose the network through network analytics to OTT players. So basically on the death scenario, signaling traffic is even growing, going to grow much, much, much faster. So what does that mean? That means my future core network must be designed for massive signaling, low latency, and data demands. So if you accept all that, then what we need to do, we need a service-enabled network that's to bring the top line up, and it needs to be able to support massive amount of data and massive amount of signaling. And I will tell you the current EPC is not scalable today. It's not scalable because it's a very centralized design. It's, you know, and I can't tell you the exact number, but let me tell you, suffice to cover the entire continent US, the number of gateways that you have in the network, you can count with the fingers of one hand. That means all that traffic goes back to a centralized point, both signaling and data. It is tremendous amount of signaling in the network that is not needed. So in the future, what we need is going to be a separation, a decentralization of the packet core. So basically, you will have functions that are centralized, and we will need to push some function towards the edge, closer to the edge, closer to the radio network. Which function we push? Where? That is an open research question. Here when we, I delineate a separation called the micro cloud. Micro cloud we think of is about the coverage of a small community. All right. So now we need to decentralize the EPC. We created these micro cloud EPCs that are much closer to the ENOBs or much closer to the access point. We will use virtualized computer service for the session controls, and we will have a centralized switch controller that are geographically distributed. So basically, we go decentralize the EPC. We will have multiple controller that will transport radio and EPC, for transport radio and EPC controllers. All right. I will say the industry is already moving very strongly towards this area. OP NFV, which has been announced in the September 30th. It is there to create a carrier grade integrated open source reference platform to advance the evolution of NFV. The most important part of that work happens to be an integration project. The first release of OP NFV will take in upstream projects, code from upstream projects, and may actually have some internal group code to put it all together to create the first release of OP NFV. So this integration project is to create an initial environment for development by integrating upstream project, create an automatic build and test process for continuous integration work so that you can give us the first OP NFV release. We're talking about the first release being in March next year, and I'm here to say Huawei is very committed to the success of OP NFV and will commit all the resource that's necessary to get it off the ground and get it going. We will contribute to the integration effort because we think it's a most important aspect of the work right now. I'm committing 10 FTE to that. We will also build a surface orientated NFV lab that is globally connected. The reason it's globally connected is this is a global network. This is the first time we have tried this on a global network. So we need a global connected testing facility for the community. We will do this. We will build it. And we'll also heavily engage in future development. Now, my one slide to wave the Huawei flag a little. I have to do that because they pay my salary and pay for my travel. And so here it is. We're committed to the NFV and indeed Huawei already has a solution. The cloud solution, which is the first step of NFV realization of NFV is the virtualization of the existing EPC. It has not. And I will stress that it has not gotten the architecture changes that I talked about previously. But this is the very first step. This you would have seen at Mobile World Congress earlier this year. And we have been testing this with over 20 tier one network operators in the world. I can't tell you who they are because they bound me to non-disclosure agreement. But what I want to say is we are very committed and we are ready to take that first step. All right. So basically in conclusion, we think that NFV and SDN are technology that will change completely change the landscape. But I want to say it is not a destination. NFV and SDN are just technologies. The destination is the re-architecture of the EPC that has not changed since GSM days. That's more than 20 years ago. So this is the biggest thing that hit that mobile call in 20 years. And it's happening at a time where we're redesigning 5G. So it's the opportune time for us to change. If we were not doing this, the timing wouldn't be as bright as it is today. So it's occurring at the right time. And we're going to re-architect that network and we'll need you guys to help because there's a lot of open questions that we have not yet answered, especially what functions and where to put those functions, where to distribute them is still an open question. But what we do know is the network needs to be more nimble, more flexible, and more scalable. All right, and we're already moving in this direction very strongly in the industry by OPNFV and Huawei is committed to OPNFV success. And that is the end. And I have 13 seconds left. Thank you very much.