So in this section, let's talk about the various deployment options for SD-WAN. There are at least three models that have emerged, and I'll seek to describe those here. Option one could be described as a hybrid model where SD-WAN is deployed on top of new hardware, but deployed behind existing network hardware on a customer premise. Now, this may be done to provide seamless overlay or, more likely, to accommodate some legacy access circuit requirements, like E1 or T1. Perhaps the more common and more flexible option is option 2 where SD-WAN is deployed on new hardware which replaces existing network hardware. And then there is a third option where SD-WAN could be deployed as a software upgrade on existing hardware, either routers or WAN accelerators. Certain vendors are planning to offer SD-WAN capability running on existing products. Option two, I feel, probably offers the most opportunity for innovation and cost-savings as it sees SD-WAN deployed on brand new hardware and offers the ability to not only deploy an SD-WAN solution on that hardware, but to migrate many other functions required for enterprise connectivity, firewalls, WAN acceleration, intrusion protection systems, as virtual VNF software running on the same platform. So here, we look at a high-level view of the end-to-end network and seek to point out the key network locations where universal CPE and SD-WAN can be deployed and disrupt the traditional approaches of providing enterprise connectivity. The three main deployment locations are, of course, at the customer premise itself, in the provider edge-- i.e. the service provider's central office or points of presence location-- or some elements of the solution can be or will be deployed in the service provider's centralized data center. So here I'm trying to explain several options for universal or virtual CPE deployment scenarios for different types of enterprise locations, different types of requirements on the enterprise side. It really demonstrates the flexibility of options that can be enabled by using Intel architecture platforms, because using Intel architecture platforms, from the smallest to the largest case, provides a consistent environment so applications can be seamlessly ported from one location, one deployment scenario, to another. So starting at the left, at a very small site or, perhaps, a SOHO office, power, space, costs will all be very constrained, and so a very small compute platform is required. One of the options that Intel can provide in this scenario is from our Atom family of products. And we see many of our partners offering solutions based on the Atom C3000 family of CPUs to address this small enterprise or SOHO use case. In this instance, the number of users at that site, and therefore, the bandwidth requirements will be low. And on a small platform such as this, you may expect to be hosting one or two VNFs at the customer premise. That could, for example, be SD-WAN, typically, and perhaps, a VPN, a VNF. While you may be hosting a broader set of services in a centralized location, so routing, firewall, carrier-grade network address translation, these things may, if required, be served from a centralized cloud location in the service provider's network. At a very large enterprise branch location, a more powerful CPE platform, virtualized CPE platform, can be deployed. And for example, that would likely to be deployed on an Intel Xeon D-2100 CPU platform or, potentially, a server based on Intel Scalable-- Xeon Scalable CPUs. These platforms offer far more computing power and can host a broad range of connectivity VNFs concurrently, maybe up to 6 or more, and can handle the bandwidth requirements of a large enterprise site. In this scenario, many of those VNFs will be hosted on the customer premise, and perhaps only the control portion, the control plane portion of the SD-WAN solution and the WAN acceleration solution would be hosted centrally in the service provider's network. So there are many Intel technologies, such as QuickAssist, accelerators for security and storage algorithms, DPDK, Intel Run Sure technologies, that can support the high performance and efficient running of VNFs on these platforms. And one of the beauties of the range of CPU offerings that Intel has for this solution is that these technologies can be run in the very smallest scenarios right up to the very largest, again, enabling portability of solutions from across these different deployment scenarios.
