Transforming the Communication Service Provider Networks with SDN and NFV
The Communication Service Provider (CSP) industry is a collection of interconnected network operators totaling over 350 operators that literally span the globe. This unified communications network is under ever increasing pressure as the ARPU remains flat or even decreases while the traffic flowing through the network is ever increasing, thus driving the ARPB (Average Revenue Per Bit) lower and lower. What is an operator to do? Part of the answer comes from observing the source of this problem, the OTT or preferably the Through The Network (TTN) cloud and media providers, they offer some glimmer of hope. This hope comes in the form of the NFV and SDN revolution underway today by the leading CSP’s as they transform to a CoDevOps (DevOps tuned to CSPs) operating model. This transformation is becoming a disruptive force within the community.
From time to time during conversations with senior leadership around a new disruptive technology, we find that the new vocabulary can introduce some confusion. Such has been the case around SDN and NFV. While these terms are not interchangeable their practical application often contain elements of both NFV and SDN, hence the confusion or mis-speaking that occurs.
Today, hardly a week passes without a significant announcement of a collaboration between a CSP and an SD-WAN solution provider
The guidance that has helped navigate these conversations is to describe SDN and NFV as two different dimensions like the x and y axis from the earliest days of our formative years studying (what for some was the dreaded) geometry. We also need to introduce the z axis to complete the analogy. Interesting applications reside in the three spaces defined by these three orthogonal vectors. If we let the SDN axis represent automated configuration, policy and control; the NFV axis represents the cloud native nature of the software application or processes, and the z axis is the product or integrated functionality of multiple applications then we begin to understand the relationship more clearly. The business realizes benefit from the object with volume that resides in this three dimensional space. Just like our analogy implies, only considering one dimension leaves us with a flat line or if in two spaces, a simple plane that lacks volume.
CSP networks of the past have been built with ‘purpose built’ platforms and solutions, while these have volume in the operational space of a CSP, they are static and are difficult to configure, maintain and operate. This scale problem places the CSPs at a significant disadvantage when the network expansion and operating costs are linear and the traffic rate is growing non-linearly. Applying the Cloud like CoDevOps model to a network constructed with Standard High Volume Servers (SHVS) offers the CSP’s the transformative model to continue to compete and remain profitable against the ever increasing workloads flowing through their networks.
Let’s consider the workload on the CSP network vs. their internal business operations network. Today this network loads of 90 percent video traffic are not uncommon, and in some cases in the high 90 percent range, various web and IP traffic, and digital voice traffic. Each of these traffic loads have different demands due to the nature of the traffic. We are all aware of the low latency along with low (relatively speaking) capacity demands of voice on a network. This ‘real time’ traffic cannot be buffered, nor can it be delayed much more than a few hundred milliseconds end to end, it is also ‘full duplex’. Streaming video requirements that differ from those needed for transporting voice, the first is the massive bandwidth demand this places on the network, and the opportunity to compress into different formats for different display resolutions, and in some cases pre-staging near the consumption point for pre-recorded streams, e.g., non-live video. Rarely is this video content full duplex, rather a massive flow in one direction and a limited flow in the reverse direction (viewer back to source). Additionally, the VR and AR workloads must pass through the same network, and yet be treated differently. A statically configured network where capacity is designed to match the workload leaves resources stranded.
The CSP reduces their OpEx through the automation afforded by SDN, fewer high-cost certified networks engineers are required to maintain and operate the network, and NFV reduces the CapEx when workloads are deployed on SHVS over purpose built solutions of the past. While savings are attractive business drivers, even better business drivers are those that come about in the form of new and possibly innovative services.
The end user benefits from a more agile and possibly a more rapid ability to access features through service portals from their WAN providers. The programmable network provided by a U.S. Tier 1 CSP is going on two years in the market and could now be considered a main stream product. With this service, an enterprise customer can dynamically alter the policy of their WAN in near real time, an activity that in the past could easily have taken months. A more advanced application that is now emerging is the Software Defined-Wide Area Network (the SD-WAN). Today, hardly a week passes without a significant announcement of a collaboration between a CSP and an SD-WAN solution provider. Interestingly, this use case has the potential to disrupt the MPLS revenue of the CSP. The SD-WAN enables the enterprise that has branch and central data centers to allow non-critical traffic to be removed (e.g., not tagged) from the MPLS stream and flow directly from the branch to the Internet without being transported to a central data center. This can occur on the same physical interface, or even in some cases over different interfaces. This last use case introduces two additional possibilities, the first is for the enterprise to measure in real time the QoS and if necessary, make routing policy changes in real time. Second, for large enterprises to have multiple service providers WAN interfaces available and do quality, time of day, cost or make any number of other policy decisions (possibly driven by machine learning), in real time and under their own control.
Its more than just the provider edge applications, this transformation reaches deep into the core of the network, affecting the EPC, IMS, GiLAN and nearly every other element of the network. The good news is we are only getting started—5G is already heating up for select CSPs, and it is clear that 5G will be built on SDN and NFV. Both the vendor ecosystem and the operations teams within the CSP are going to be challenged to meet the demands of 5G. The use cases will bring to the network a new level of scale, and place additional pressure on the infrastructure costs. The expected growth of connected devices and the incredible levels of traffic they generate, will continue to demand that new and cost effective solutions be driven into the network.
While we may not accept that the new language has reached the level of common use that it no longer has special meaning, what we should agree on is that the transformation of the CSP network is well underway, and SDN and NFV are the underlying concepts at the foundation of this transformation.