When you talk a little bit about the Intel technologies at this point that support this, I like to say that CPE is different. What I mean by that is, if we look at the core of the network, when we talk about how we're going to build and manage and operate the core of the network, the functionality can be very much standardized operationally as we look at those elements that go in the core of the network, except for CPE, CPE comes different. One of the things that's different about CPE, even if it were distributed, even if we take part of the firewall functionality and put it into the core of the network for banker or the barber shop, we wouldn't want that firewall necessarily to business decision, for example for a financial institution, they're going to want to be able to control that at their brick and mortar location. So, they're secure that that data's also secure by a firewall that's physically at that location. They may have regulatory compliance that drive them to do that as well. The other part of it is important to understand is that in CPE, there's always going to be an element where the enterprise itself is paying for the electricity, there's going to be a device of some elk, that's at the enterprise location that provides that interface, that termination, that connectivity, whether it's wired or wire-lined or RF over some type of a fixed 5G access, or maybe even a 4G access, is it that enterprise itself is going to have a device that provides that interface between their local area network interfaces and those access points added to the common service provider network. Because of that, that enterprise itself is going to size those elements, so we talk about sometimes t-shirt sizing, so you think about t-shirt: small, medium or large. Even in the enterprise CPE with their functionalities, if you've got a small branch office as opposed to a very large branch, here, you've got someone that's got four-five users at a branch office and are not using a lot of video production generation but then you've got another corporate location that's got up to a thousand users, their need for access to the amount of capacity that are going to generate in that network is going to be different and the size of those of platforms. They're saying they're consuming electricity at the enterprise location itself, should be sized appropriate for that, so that they're not over-subscribing it. So, when you look at technology from a two core, to a four core to an eight, the 16 core type of a device, it scales up linearly, the larger enterprises are going to use them to large multi-core devices and at the lower end we may have a more cost-effective solution in that product. Nevertheless, at the middle layer, these are still going to be virtualized-type platforms, so that then the VI layer can be common across all this and there's no recompilation, the same applications that are running on the two-core device can run seamlessly and identically on in the 16 or 32-cores type size devices that will exist out there. I don't need to change them, simply I can apply more application resources to them so they support that increased data need that comes into play. So, I talked about the barbershop and the banker as being different, and because of that, we actually see that inside the SD-WAN Deployment we've got different models, here they're identified as three different options one, two and three. A hybrid model where the hardware is deployed behind existing network hardware and then you have legacy interfaces they still exist, these things take a long time to go away whether those the E1 or T1 based on the European or the ANSI standards whether these are DS3, is there some type of an OC3 type of an interface. Then there's a new model that replaces that equipment with a more tightly integrated SD-WAN and a CPE environment and hopefully those are non-legacy interfaces. In some cases they will be and in those environments you may have to look at an adapter device, that provides inter-connectivity between that legacy interface T1 human type of interfaces and a more IP-centric type of interface. Then finally, the last option is more of a distributed architecture, where we've decided that we're going to split some of that functionality across there and still manage those elements through an SD-WAN, where portions of that, you can do some of the firewall may run deeper in the cloud and others and SD-WAN and can be used as a stitching function to bring all of this together. So, overlaying of the SD-WAN is a particular example, so the head end controller itself that control plane device that needs to be aware of that policy, of those layer two and three type of applications again common layer one, in some cases are different layer one. But SD-WAN control has to be able to detect the nature of those packets and identify them either based on information at layer two or layer three and then control that xLAN LAN connectivity. The other use is, when you've got congestion in the network, you may be able to take the opportunity, if you will, to move traffic from the IP network over into the MPLS network and depending on the traffic, you may be able to move it in the other direction and in that case, you may have to apply some additional security encrypted differently as you move things over the broadband, so there's a redundancy issue that SD-WAN can introduce that you may not have if you've got only a single interface that's available there. Then obviously in the office, using VxLAN type capabilities in the overlay are possible to be introduced, this was sort of implied before tunneling through the IPsect. There are a number of vendors, both pure software vendors that are out there as well as those that provide a complete integrated solution that is the hardware, the NFVI and a collection of those service-chained functions on top of them together here, so here you see some of the use cases for either SD-WAN and some of the OEM/ ISVs that are available to it. Same thing in the security, routing and the WAN optimization and you can certainly contact any of them directly or check out Intel network builders for additional information on any of these opportunities.
