“Network Function Virtualisation (NFV) will represent a 5 to 10-year evolution cycle for communication service providers (CSPs) and there is a broad consensus that NFV is needed and is going to happen,” Javier Benitez, senior network architect at Colt Technology Services tells VanillaPlus. “However, at the same time everybody acknowledges the huge transformation that will be needed.”
So, in search of a few ground rules for NFV deployments, Jeremy Cowan asked several industry experts what to do and, just as importantly, what not to do when migrating a CSP to an NFV-based architecture. The following experts offered us 20 tips:
Javier Benitez, senior network architect at Colt Technology Services
Béatrice Piquer-Durand, VP at Ipanema Technologies
Ben Parker, principal technologist, Guavus
Gary McKenzie, senior architect, Peer 1 Hosting
Gerry Donohoe, director of Solutions Engineering, Openet
Shaul Rozen, director of Product Strategy, Amdocs
Do
- Do identify what the key NFV user cases are that you are going to focus on, from a CSP business point of view (backed up by the relevant business case) and a technical point of view (based on current product’s needs).
- Avoid the temptation of developing NFV without a proper Orchestration layer that is fully integrated with your existing operations and business support systems (OSS/BSS). The days are long gone where network functionality per se was the most important aspect and the management of the solution came as a second priority. NFV can’t be successful if orchestration is not there from Day One.
- Do avoid running proofs of concept (PoCs) forever. If you have identified a key NFV user case, try to bring it into production, as this is the only way to cover all areas, aspects and teams impacted by this new technology.
Javier Benitez, Colt Technology Services
- Do make sure you have sufficient knowledge of the underlying platform on which you’ll be running your NFV solution. Whether it’s VMware, Openstack or something else, you need to ensure your foundations are robust and reliable, that’s not a skill which always naturally sits in network organisations so make sure you’re engaging the people you need.
- Do understand the way NFV changes your sizing and performance requirements, and pay attention to which NFV solutions run better on which platforms (VMware still often outperforms KVM-based virtualisation we find, though the gap is narrowing).
Gary McKenzie, Peer 1 Hosting
- When it comes to deciding which services should take priority in this drive to virtualisation do consider:
– The value of the virtualised services to both the end customer and the CSP, and
– The ability to seamlessly orchestrate not only the deployment and configuration of the service in the virtual environment, but its ongoing operation and updates.
High value services that can easily be rolled out and orchestrated globally will be more appealing both to end customers and the CSPs themselves. Services that require a lot of bespoke configuration point-by-point will offer a much lower return and higher overhead to administer and will be much less appealing. Global application-level services such as WAN optimisation and firewalling are ideal first targets for this reason.
Béatrice Piquer-Durand, Ipanema Technologies
- Most carriers are thinking about NFV from the perspective of the benefits of virtualisation. However, there’s much more to it than that. Do think, for example, about whether services should be decomposed; meaning would an operator wish to break out the components with
in a virtual network function so that multiple vendors can participate? A good example of this could be a packet data network gateway (PGW) where Cisco terminates the GPRS Tunneling Protocol (GPT) Session,Juniper performs Network Address Translation (NAT), and Alcatel-Lucent performs the Dynamic Host Configuration Protocol (DHCP) function. - Embed analytics to perform anomaly detection and feed critical information to orchestration systems. This will improve reliability and the operability of virtual networks. By having a third-party solution monitor health, performance, KPIs, and other stream data, it is possible to improve reliability, scale, and performance of the virtual network.
Ben Parker, Guavus
- Do adopt a vendor ecosystem that is open and embraces standardisation. One of the main drivers of NFV is to reduce vendor lock-in and use of proprietary systems to enable business agility and reduce costs. CSPs need to be wary of vendors that promise advances at the expense of open and standard-based solutions.
Gerry Donohoe, Openet
- NFV brings IT and cloud technology and operational principles into the network. Create cross-organisational Network-IT teams that bring their combined expertise to the table and create new value and understanding.
- Some of the key values for NFV are agility and flexibility. Process and action automation across technologies and data centres is crucial to realising these benefits. Bring these topics into the discussion early on.
- Get your internal business partners excited about the new flexibility that NFV will bring to product management. NFV allows service providers to better serve niche segments, test new offers in the market or create innovative flexible business models.
Shaul Rozen, Amdocs
Don’t
- Don’t aim for a full transformation at once using NFV. Focus instead only on the agreed priority use cases. In the initial phases NFV brings a level of complexity that CSPs need to be very careful in managing, especially during the learning curve, while protecting the existing business.
- Don’t wait until NFV is fully standardised or GA products are available. NFV (and its brother,
software defined networks, SDN) move CSPs to a new world ruled by software where standardisation is replaced in many instances by OpenSource and community development. If you want to influence or have your say in this new world you need to start playing with the technology as soon as possible and be active in sharing your views (even your code!) with the community. - IT and Network teams need to work together on NFV from Day One. NFV means network services in the cloud, so there is no way NFV can be successful if either part is missing.
Javier Benitez, Colt Technology Services
- Don’t rush into it assuming it’s just a virtual firewall / load balancer / other and it’ll be pretty much like a physical one. When you’re deploying NFV at scale in production environments it has its own unique set of challenges to master.
- Don’t treat NFV as nothing more than a replacement for the physical product, the abilities it brings to the table (adding a new network feature into a solution in less than five minutes) are new and should be treated as such. The old business models won’t necessarily fit any more.
Gary McKenzie, Peer 1 Hosting
- Don’t forget that while initial trials may be isolated, NFV is a transformation that affects operational processes, business processes and even organisational change.
- Don’t limit your NFV thinking process to the network functions level, take the discussion to a higher level and think of NFV in the service level. This will make it more relevant to end users and the value you would like to achieve.
- Don’t forget the end-to-end vision as you start implementing NFV in different areas of the network. If you do, you might end up with the same silo network, with multiple management systems, just virtualised.
Shaul Rozen, Amdocs
Comment here or on Twitter: @jcvplus OR @vanillaplusmag
