 Hi, I'm John Schimpf. I'm Director of Business Development for the Wireless Business Unit inside Cavium. Who are you? Hi, nice to meet you. I'm Kenji Blue. I'm in Cavium, Director of Solutions Architecture. And right here you're talking about running M-Cord live demonstration. What do we have in here? There's four blades. So what we're showing here is the virtualization of the network core and the baseband unit running on 64-bit ARM-based, Thunderext server blades. Let's go over there. You have some slideshow showing more information about, you're talking about M-Cord on ARM, right? So this is, what is M-Cord? M-Cord stands for the mobile central office re-architected as data center. So is this have to do with virtualizing all the stuff about a server? What is it? This particular cord implementation is about virtualizing the telco network. So the telco cloud servicing customers using a ARM-based architecture. And there's a bunch of companies working together on this, right? That's right. So the companies that are involved with this program include network operators, software companies, silicon companies like Cavium supporting the ARM architecture. It's got to do with NFV, SDN, all that stuff. All that stuff. That's right. So the ability to use general purpose servers to perform all of those network functions. So is it working right now? Does it work on the Cavium hardware? Absolutely. Thunderext 1? Yeah, we are doing live demo. And the rack of CPU processors, general purpose processors, you see are based on the Thunderext 1. And obviously we have announced our second generation as well. But even the first generation, we have 96 cores on one server blade. Because each processor has 48 cores and is a two socket configuration. Thunderext 2, how soon is it coming? Did you say yet? It's coming soon, right? Yeah, it's announced. But you know, we can focus on the current demonstration because it is very interesting. As you mentioned, it has both NFV, SDN, and there's also multi-access edge computing, which is a very new technology. Because as we've seen, there's a lot of devices. There are new mobile technologies. So at the edge of the network, you can do a lot more services. And that's part of our demo as well. So traditionally, they would have a lot of different things in the server area, what do you call it? They would have lots of different things they want to replace with just one KVM rack, or how is it going to work? Well, it's not replacing things with a KVM rack per se. The idea is to have a scalable way to do all of the network element processing. And the case right now is for the usual network operation is to have dedicated appliances. So switches, routers, gateways, all of those pieces of equipment are dedicated to doing their particular function. The idea with this approach is to use general purpose servers running application software being managed by a series of virtualization techniques that allow those applications to come into play or to scale. If the requirements on the user side increase, all you have to do is allocate another server. You don't need to allocate another piece of dedicated hardware equipment. When that piece of equipment or that server is no longer needed, it can be dedicated to another use case. So it provides a lot of flexibility and helps to reduce the cost and the power consumption within a data center for supporting the telco cloud. And it's important to support peak times. You don't want to have too much stuff you don't use, right? In particular, that is one of the cases that the operators are trying to solve for. So if you're using general purpose servers, again they can be used for multiple different reasons during the course of a total day. But they're available to support those peak times when dedicated equipment is required. On the down peak portion of that, they're sitting idle. And the expense is still there for all that equipment, but they're not really being used during those times. And this is called the F-Cord on ARM, and I guess it also works on Intel. So how good does it work on your ARM right now? So basically with ARM, it provides a much bigger ecosystem. So customers have a choice of both x86 and ARM servers. And obviously the diversity of hardware is very important, because as we get into the edge computing and IoT computing, you could have cases where you need a very low cost server, cases where you need a mainstream server. And basically it's all about having better choice and better ecosystem for the entire industry. And do you already claim to be better than Intel in terms of cost value and stuff? The Thunder X1 is already better? Yeah, Thunder X1 is much better in terms of cost. I mean, you can see some web hosting. Now there's like a bare model hosting company where you can see the pricing. If you want to rent a VM per hour, there's a cost for the Xeon, a cost for the Thunder X, a cost for the Atom, and you can see that range. And definitely the Thunder X provides a very good value there. The best value, right? Absolutely. The best performance and value combination. And then with Thunder X2, it's going to be even more better value? Yes, yes. And that's coming. So, and then there's all these apps, are those apps that runs in this system? Yeah, so this is a graphic GUI, basically for the Cod project. So the Cod project is about, as John mentioned, the central office. We architected it as data center. So in the central office today, there's a lot of legacy appliances, which only does one thing. That is built for. So with virtualization, we are using ARM servers and running virtualized version of these applications on ARM servers. And in this particular demo, we are running the virtualized base station, the base band processing, virtualized mobile network. So those are all running on virtual machines, on open stack as well. And in addition, we have a partner trend micro with security services running also as a virtual machine on the same infrastructure. So they're running like an antivirus? So interesting, there are multiple applications in this suite. These applications provide, for example, deep packet inspection for malicious attacks and malware. It also does intrusion prevention. It does parental control, also URL filtering. So for example, a parent can block his child or her child from going to certain websites or a company can apply policy. For example, a company can say if you are working and you're using our network, you cannot go to these websites during work hours or something like that. So it's a very flexible policy. But at KVM, you have different ThunderX. You have one optimized for one thing for the other. So you wouldn't buy just the standard one. You would still buy different ones for different news. So there are many kind of different servers or other applications in the industry, right? So from a mainstream server perspective, we have ThunderX as a processor. And it does have a lot of different form factors. You know, a typical 1U rack server. You can have high density OCP versions, telecom OCP versions. So those are all available from our partners like ODMs and OEMs. And also with the same technology, we have implemented a smaller version, which we call Oction TX family. So in our mainstream, this one has one of Oction TX processors. It's an industry-grade IoT gateway. So you can see it's very rugged design. It's very stable and it's very adaptive to any kind of harsh environment. So it has a lot of capability of connecting to a lot of devices locally as an IoT gateway and a server. And it will connect through analysis and connect to the network. And that would be for 5G future? You would have lots of base stations running like stuff like this? Yeah, so this is actually a little bit different from a base station. So this product basically works with a base station by connecting to a lot of IoT devices. It can do analysis, protocol conversion of these data collected from the devices and then talk to, you know, if it's a wireless, could talk to a base station and get the data up. Or it can go through some wired connection and get the data up. So yeah, key thing is it can do a lot of intelligence analysis locally without having to send all the data up to the cloud. So basically, is it kind of like an app store for the telcos? Absolutely, that's the idea is that as their customer you can choose different kind of networking applications from different vendors and it's a bigger ecosystem, more flexibility. And anybody can just add some apps. Absolutely. And they might be able to sell them to some telcos. Absolutely. Yeah, absolutely. Yeah, and interesting thing also is that, you know, we talk about both X86 and ARM being a server basically more choices to the consumers. And in this case, consumer could be service provider, enterprise, etc. And they all use the same industry standard platforms, you know, like Linux kernels, OpenStack, KVM, containers and also, you know, SDN controllers. These are all industry standard software platforms for virtualization, for NFV. They all run on today. And this is very important for everybody who is watching this video because everybody wants to have Gigabit LTE, 5G, everybody wants to stream 4K videos today. 4K, IHDR phones, they want to upload 4K, they want to do all these things. It's impossible if you're not doing what you're doing right now. That's right. So basically, you need a lot more hardware, a lot more cost effective hardware to do all of this and also relevant acceleration to do all this as well. So you're enabling basically a better experience for every consumer, like every end user at the end. Yeah. Because there's a limit to how much you can do with super powerful, what's called power-consuming Intel. You need to have more efficiency. Absolutely. I mean, like a lot of these applications, the reason why they are on purpose built appliances today is because it does need a lot of unique operations. For example, on an LTE network, between your user device, which could be a cell phone or tablet, to a base station, all the traffic is encrypted by algorithms which are unique to the LTE standard, like Snow 3G, ZOOC. These are not algorithms that are used other places. And as a general purpose server, you will not get hardware acceleration for these algorithms. But with a choice of more hardware, then you can have acceleration based on standard APIs. So the standard software with a diversity of more optimized hardware, you can get better performance and value and lower cost. All right. So this is happening right now. Is there any announcement about how successful the ARM server is thus far? Or are you mostly talking about the future still? Yes. So ARM servers are being deployed. As I mentioned, there are web hosting companies. You can rent cycles based on a dual socket 96 core per blade ARM servers in a web hosting manner. So it's very cost effective to try and test your software. They're also in terms of other infrastructure networking devices, security devices where ARM is used as well. So ARM is not just for the handheld. It's been used in the infrastructure in a successful manner. And is everything just working out as it should? Or is it possible that there might be some apps that don't work yet to enable? Like you said, there was specific requirements for LTE encryption, all that stuff. Is everything going to work on this kind of system? Yes. And all the x86 stuff is going to work on ARM? Is it already working? Yeah, most of them are already working. So we have to set the standard platforms and that's the important first step, right? Because you need the virtualization platform, kernels, OpenStack, SDN controllers. So those are all working. So the ecosystem can use that to build upon more and more applications.