Key Challenges to NFV Implementation in 2020

As we face the world's first truly global crisis, the data networks we have spent the last half-century developing have never been more needed. Our 3G, 4G LTE networks, and even our 5G networks are in the midst of a planet-wide trial-by-fire.

In our current phase of social distancing and teleworking, tens of millions of individuals now rely on mobile and wired networks to deliver information and communications at rapid speed. Network functions virtualization (NFV) has turned out to be essential to delivering the agility and flexibility everyone is increasingly demanding from their connectivity services. Many large carriers have come out and attributed their agility to scale to their recent NFV and virtualization initiatives.

Fortunately, communications services providers (CSPs) had started the process of shifting to virtualized and cloud-native architectures. Many have become providers of technologies like SD-WAN, and leveraging containers to provide OTT-like service delivery over their networks. 

At the same time, such an epochal shift is not without its challenges. Some NFV barriers are related to industry-wide inertia — latency of the human thought, if you will. Others are related to the specific and unique circumstances that NFV introduces.

With these forces in mind, there are two categories of general NFV-related challenges that CSPs face:

1. HR-based NFV challenges, including a lack of needed talent and the need for training/re-training current personnel
2. Vendor-based NFV challenges, which encompasses the extremely fragmented solutions ecosystem, a lack of standardized NFV infrastructure (NFVI), and a dearth of collaboration outside of specific licensee relationships.

Additionally, certain challenges come amidst an emergent opportunity: the value open-source communities provide in NFV development, beta testing, and use-case-defined implementation.

While all of these NFV challenges are significant, they are not insurmountable. As CSPs continue to deliver true 5G services and stride towards NFV and CNF, the solutions to their ongoing challenges provide the seeds for an even newer generation of technologies for tomorrow — and beyond.

Industry-Wide Upskilling and an Intense Competition for Talent

Many of the advantages that NFV provides come specifically from the technology's abstracted nature. Moving from hardware to software-based solutions uncouples services from the constraints of physical technology, but at the same time it puts people in unfamiliar territory.

Telecoms professionals who began the ascent to their current roles in the 80s, 90s and even early 2000s tended to be highly "device focused." New advances in services came directly from advances in server/router architecture as well as other evolving hardware. Thereby, understanding the hardware was necessary to understanding the services.

Coming to grips with your typical virtual network function (VNF) involves unlearning many of the hardware-dependent skills that once helped individuals advance in their field and innovate new ways of delivering services. NFV and the subsequent cloud-native architecture involves not just new technologies but entirely new, completely abstracted ways of thinking. That includes new ways of thinking not just about services and infrastructure but also the way data is managed across highly containerized, cloud-hosted environments. Entirely new competencies are needed.

To show how behind we are in the CSP world, we refer to a 2008 Networking World article that describes the intense competition seen in enterprise IT recruiting, with one interviewee predicting: "The skill set required for this technology is only going to increase, as is the number of mission-critical applications hosted on a virtual architecture." 12 years ago, enterprises were in the throes of VM transition, and 8 years ago, with the initial introduction of NFV, the CSP world tried to initiate the same in their production networks. Today, it’s no longer about simple virtualization, but also a move to cloud-native platforms.

CSPs need to catch up — and compete! — with these other industries. NFV and cloud-native transformation will give rise to new specializations and the need to collaborate with other domain subject matter experts.

Here are some of the most salient human-resource related observations in our most recent canvas of leading CSPs worldwide:

• "AT&T realized that being successful in changing from a hardware company to a software company would require different skills and significant retraining within their existing workforce."
• "Colt’s skills development strategy is built around two axes. One is to promote internal transformation by defining the right organization and bringing the right profiles and skills in joint teams, e.g., network and IT profiles, while defining upskill programs where required. The second strategy is to acquire new talent in areas like software development to help deliver the on-demand customer experience."
• "On the topic of virtualization skills and talent, Deutsche Telekom is in competition with the webscale giants — Amazon, Facebook, Google, Microsoft, Netflix, etc. — for people with the necessary software skills. At the same time, there’s a focus on retraining its own workforce: a program successful with most people, but not all. Deutsche Telekom has also brokered partnerships with various universities as a pipeline for new talent."
• "Telefónica has discussed a plan to retrain 6,000 employees in areas such as security, robotization, analytics, web development, business consulting, IT, and agile methodology capabilities. It has planned to double its training budget and increase the training hours per employee by 40%. This overhaul was necessary because Telefónica still employed too many analog workers in a digitizing world."

Frustrations Around a Highly Fragmented Vendor Ecosystem

Generally speaking, there is limited availability of proven, functional, high-performance cloud-native products offered by vendors that CSPs can leverage to accomplish their goals.

Vendors have come to play an increasingly large role in enterprise productivity, which now encompasses the delivery of highly innovative — and, sometimes, experimental — solutions. 

The pressure of this role means that many enormous enterprise projects hinge upon whether or not upstart companies can provide all of the promised deliverables. Even when the deliverables fully meet expectations, CSPs must still contend with the challenge of integrating all of the various vendor-based solutions into a cohesive whole. Most CSPs report significant challenges in integrating, troubleshooting, and deploying multi-vendor solutions. 

VNF and other NFV service vendors are being highly innovative, but they also have little motivation to commoditize their key revenue-generating services. There is also little coordination, meaning that industry-wide standards must be developed purposefully and then enforced upon them, rather than CSPs holding out hope that a standard will emerge all on its own.

On top of these challenges, the general immaturity of modern vendor-based solutions makes itself painfully known. Pay-as-you-go licensing models and bring-your-own-license (BYOL) are still "working out the kinks," and there's little consistency from vendor to vendor. 

Speaking to these challenges, a few major CSPs had the following observations:

• ChinaMobile has, "concerns around system-level performance differences between different infrastructure vendors as well limitations in service assurance."
• "The central challenge Colt faced in the move to virtualization was the disaggregation of hardware from software, which placed the integration burden on the company despite support from their vendors. A key lesson learned here is that nothing is plug-and-play, despite claims made by vendors."
• "Colt’s biggest expectation for its software vendors is to align with the industry VNF onboarding standards (ETSI/ONAP) so that the time and cost of onboarding new VNF vendors and applications can be dramatically reduced. The current industry status, with long onboarding times and high support costs, will not be sustainable in the future. [...]The ongoing transition in the telecom industry towards a new software world is demanding more and more support from external parties and system integrators."
• "Whether the vendor is a cloud provider or network solution provider, Deutsche Telekom needs high quality from its vendors so it doesn’t end up as a beta tester."
• "A central challenge faced by Orange has been that different operating units are running different variants of NFVI. This adds complexity for Orange’s selected VNF vendors, who need to test software against each organization’s unique NFVI. If it proves incompatible, modifications could be needed just to serve the needs of one particular team."
• "In a joint white paper published with Analysys Mason, Telefónica publicly stated its concerns about vendors’ OpenStack implementations. Specifically, both companies felt that the vendors’ cloud technologies did not support the specific performance and distribution requirements of network functions. They stressed that they would continue to collaborate with multiple technology partners and influential industry organizations to ensure their requirements were reflected in the appropriate roadmaps."
• "Being early to market with virtualized services, Verizon faced the challenge of explaining its strategies to vendors, for some of whom this was their first NFV experience. In doing so, Verizon became its own system integrator, which represented a challenge for several internal teams. Unable to use existing internal tools for operations and services, those teams needed to create new solutions. Disaggregation presented several challenges including optimization around technologies such as DPDK and the test environment for the transition from physical network functions to VNFs."
• "Vodafone recognized an industry-wide challenge caused by each CSP selecting different variants of NFVI. VNF vendors need to test their software against each CSP’s NFVI platform. Even within Vodafone, there is the possibility of a different NFVI configuration for every VNF deployment. This would result in a multi-silo deployment in their networks, which defeats the purpose of moving to cloud and leveraging a common, shared infrastructure."
• Learn more about NFV development, along with the observations of the world's biggest CSPs, in our recent 2020 State of NFV Report.

Challenges and Opportunities within Open Source Development

One surprising development occurring during the development and implementation of NFV solutions is that many major CSPs are turning to collaboration with open source communities. This allows them to contribute to development while joining forces with impassioned volunteers hungry to make untested concepts a reality.

AT&T, for instance, states that 100% of its NFVI software and 40% of its internally developed software is based upon open-source contributions. 

"AT&T leverages open source in order to work together as a community to innovate and deploy faster, without starting from scratch every time," they report, adding that they believe, "this allows their teams to focus on building new technology instead of reinventing technologies others have already built."

While collaborating with open source projects has its advantages, it also has its drawbacks. For one, AT&T cites a difficulty with, "sustaining open source communities once the hype has died down and projects mature."

China Mobile, similarly, "found wide variations in system-level performance as well as in high availability support beyond the scope of standard open-source distributions."

As for Colt, they believe, "there should be much more harmonization and rationalization in terms of open-source initiatives as well as standardization efforts. The industry needs to merge and combine some of the overlapping existing initiatives in order to foster de-facto industry standards," adding that they, "would like to see the industry to focus on initiatives to simplify the overall NFV architecture, Lean NFV being an example."

Challenges of Today Invite the New Solutions of Tomorrow

Challenges invite ingenuity and encourage market growth as more people embrace NFV. 

Vendor fragmentation, for example, will face pressures that will inexorably lead to the adoption of one (or a few) main architectural standards, allowing for greater "plug and play" compatibility and less work within integrating various solutions.

The need for talent and subject matter competency will encourage entirely new areas of study, creating a talent pipeline that extends through universities, and possibly even secondary schools.

Gradual migration towards virtualization will inevitably reveal inertia, but competency will come with time and familiarity.

The growth of NFV is already underway. While some barriers may slow progress to the eventual destination of full virtualization and cloud-native architectures, successes by several CSPs prove that these obstacles are not insurmountable.

Overall, the benefits of virtualization are too great and the dream is too big to let anything stand in the way. Every step of progress brings us further towards a future of data networks ripe with more capability and more potential than ever before.

Learn more about NFV development, along with the observations of the world's biggest CSPs, in our recent 2020 State of NFV Report.