NFV strategy, from trial to commercial operation
Although NFV and SDN have been discussed since 2012, most operators have several years left to come to a state where most of the commercial services are run on a virtualized infrastructure. The majority have started in some fashion and fairly common are trials or PoCs (proof of concepts) where a single application, such as vEPC or vIMS, is running on an appliance type of NFVI. Very few are running a large-scale commercial telco cloud to cater to a multitude of VNFs from many vendors.
However, we now see a sense of urgency to get something really working, because eventually this will be needed for many 5G features as well as for IoT services already on 4G. There are operational benefits already today and being late to the party can be a risk.
There are several reasons for being late, such as the overall complexity and the challenges of integrating all the components; the use of open source; the organizational changes needed due to the lack of NFV competence; and the fear of making the wrong choices, which can turn out to be costly in time, capex, and opex. How should you take this on? Ericsson has been working with several leading service providers for several years, and we have created a series of articles or eBriefs to share some of our insights on the NFV transformation journey.
What have the leading operators done?
Some operators have managed to get their networks to a stage where they are now capable of starting multi-VNF operation on a single NFVI platform with desired VNF performance. Typical for them have been the following:
- An early start with a clear vision of what to accomplish that includes an e2e view across the whole architecture and its operation
- A readiness to change the architecture and the way of working
- A realistic view on how many vendors and solutions can be mixed and matched in the beginning
- A close cooperation with some critical vendors to master the migration roadmap
I have started, but how do I finish?
The answer depends on who you are and what you mean by "finish". A first type of "finish" is to have an NFVI platform with some applications combined with an ability to orchestrate that and run the legacy network in a hybrid mode.
The first step is having a strategy and migration plan that includes services, services orchestration, network architecture, operations, security and vendor strategies. The large T1s have started early with often radical transformations, for example, AT&T with its Domain 2.0 as a blueprint. This applies equally to several global T1s, which typically have their own version of the blueprint. The strategy is to integrate themselves, choose best of breed, and avoid vendor lock-in. Most have started to build an NFVI platform and now have a few VNFs running for parts of their subscriber base.
This requires substantial investments and has so far taken a bit longer than many expected. Smaller T1s and T2s typically start with something that is more pre-integrated and pre-tested, because they don’t have the same resources as the larger players. This has proven to be one of the fastest ways to get to a situation with something working, including a change in how to organize the operation and align the IT and telecom teams.
For smaller operators, the most viable strategy will be pre-integrated appliance-based NFVI platforms combined with aaS solutions. There are solutions today for MBB+VoLTE+IoT services that can be set up in a few weeks.
Does NFV deliver on its promise of speed, agility, and no vendor lock-in?
We are starting to see some evidence of NFV delivering on its promise where lead times for upgrades have gone from weeks to minutes, services provisioning has gone from months to days, and there have been substantial TCO reductions for vEPC and vIMS.
You can discuss what we mean by vendor lock-in, but a complete ability to quickly swap suppliers will require a very standardized product such as a PC. We may not be fully there, but clearly a lot has changed now because we are starting to see many VNFs being able to run on any NFVI based on more thorough certification towards the major VIMs and hardware platforms. Also as the NFVI is not fully standardized, not all combinations work so therefore pre-integrated solutions are preferred today.
5G and NFV. What is the connection?
A full-fledged 5G system with features such as network slicing and distributed cloud with edge computing and later the 5G core will require a fully working NFV platform for its operation. 5G is often coupled with IoT and industrial internet, which in turn will need automated ordering and provisioning of services based on a new eco-system. This does not mean that it’s a good idea to wait with the NFV transformation though, for several reasons:
There are benefits in operations and agility already now without 5G.
The transformation journey is a multi-year transformation likely to take at least three years even when keeping a good pace. A late start increases the risk of being late with 5G services introductions.
The new ways of working both in terms of operations and service provisioning will require some time to mature.
Your agenda for NFV
Given the development towards virtualization and 5G, it’s evident that all operators must go through an NFV transformation. The question is how and when?
Want to know more about how to overcome NFV adoption challenges and start the journey towards an NFV network? In our new series of eBriefs we will share our learnings and insights from working with the early adopters. Read our first edition, From NFV trial to commercial operation, to learn more:
Stay tuned for the next eBrief “Follow the early adopters”.