There comes a point in every technology adoption cycle when what John Maynard Keynes described as industry’s animal spirits take a beating.
In the IT industry, the analyst firm Gartner uses an appropriately mystical expression for this moment: the trough of disillusionment. It’s the point at which easy promises have evaporated. The road ahead – the slope of enlightenment – seems full of mind-numbingly hard work.
That’s precisely where carrier virtualisation is right now. In a survey of carriers published by SDNCentral in March, only 13% described NFV as “complete and mature” or “almost mature, ready for production deployments”. By contrast, 72% described the technology as “coming to maturity, more work needed”. In recent months, commentators have described virtualisation as an “albatross”, “a giant rolling hairball”, “indecipherable”, “practically unusable”.
Sitting in his hotel in Nice reflecting on the four-day TM Forum Live event he’s just attended, Michael Howard, IHS Markit’s veteran senior research director for carrier networks, has no doubt that the industry needs to press on with virtualisation: “You gotta do it. It’s absolutely key,” he says.
Howard is not the kind to underestimate the challenges. Early in his career, he worked on ARPAnet for Santa Monica-based Systems Development Corporation, typically considered to be the world’s first software company. This year, Howard published his fifth IHS Markit survey on SDN/NFV adoption among service providers, interviewing executives from 35 operators who collectively account for over 60% of the industry’s global capex. He retains a close interest in code: he likes delving into proofs of concept at conferences because they “tell us where we are, what the state of the art is”.
This is where we’re starting to see innovative new approaches to deployment: one example at TM Forum Live was the Joint Agile Development (JAD) catalyst presented by Huawei, IBM, Infosys, Spirent Communications and Tech Mahindra. JAD lifts methodologies from DevOps in the IT domain, and repurposes them for carriers. According to Sean Yarborough, senior director of marketing for Spirent, efforts like these open the way “design, validate and deploy service enhancements dramatically faster”.
Howard spoke to Global Telecoms Business about the past, present and future of virtualisation. We discussed a mixture of NFV use cases and technology domains in which carriers are making real – albeit frequently painful – progress.
1. End-to-end orchestration
The word orchestration is frequently used in NFV (and in the cloud computing market). Typically, we understand its meaning to involve “making things work together”. But Howard is keen to stress the role of automation.
“When people talk about orchestration,” Howard says, “I always ask: are you talking about NFV orchestration, or end-to-end service orchestration?”
“The example of end-to-end orchestration I like to use,” says Howard, “is a global bank with 20 locations. Say they want to set up an IP/MPLS service to those 20 sites. Let’s also assume that the operator offers different options for firewalls, some of which are suitable for the large offices and other suited to the medium-sized offices.
“Now this is a service configuration that you, the operator, want to make available online through a customer service portal. You then want OSS to translate that into a network service. You want to minimise human intervention, because if humans get involved, like they do today, this whole process can take months. The bottom line is that you cannot deliver on-demand services with agility without a very high level of automation.”
2. Orchestrating virtualisation
The orchestration of NFV itself resides in the MANO (management and operations). In functional terms, a MANO needs to take care of three jobs: controlling compute, storage and network resources; lifecycle management and event reporting; and orchestration.
This remains one of the most fluid and hyperactive virtualisation segments.
Open source developments are continuing apace – ONAP, ETSI’s Open Source MANO, Open LSO, Open Baton, Tacker. Here, too, cloud software vendors, with their understanding of virtual machines and containerisation are very much part of the picture. Above all of this lurks the question of the optimum relationship between MANO and OSS/BSS: if one thing seems likely in 2017-2018, it’s additional successful efforts to merge and integrate the functions of both.
Howard fleshed out one of the early architectural concepts for an “orchestrator of orchestrators” to manage NFV. “When I looked at what SDN/NFV needed to look like, I placed a big cloud on top called the orchestrator of orchestrators. Applications and customers, appealed to that controlling box and it has within it was a services function but also a network function that translates that service the customer is buying into network terms. My thought was that where the OSS feeds into the orchestrator of orchestrators, they would have to merge.
“That’s what NetCracker, Amdocs, Nokia, Huawei and ZTE have all done: they have been adapting and sometimes rewriting their systems to become the orchestrator of orchestrators. There’s a lot of basic functions in today’s OSS/BSS that you still need. However, they have to be adapted to hybrid network environments: physical and virtual. You bring in all the relevant new elements.
So for example in CRM systems, you need that interface with the customer: you have to add a portal and the concept of on-demand services with durations.”
Conceptually, this makes sense. On the ground, integration between OSS/BSS and MANO remains headache inducing. Just 10% of the end-users questioned by SDNCentral said integration was “going smoothly”; 51% cited the need for “a lot of custom coding and services” and 37% agreed with the suggestion that there are “no good solutions today that integrate both well”.
3. VNF On-Boarding
The continuing discussion about VNF on-boarding within NFV underlines just how fluid (read: immature) NFV remains. The speed with which you can configure, test, start, stop and restart VNFs is a key determinant of agility. It speaks volumes that at this stage in the game, TM Forum is championing an award-winning Catalyst working group that aims to cut the time it takes for VNF on-boarding from six to eight weeks to a day.
The group is backed by AT&T, China Mobile, Orange and Verizon. Howard witnessed the Catalyst group’s work at TM Forum and liked what he saw. “Now they’re taking that concept of the package to two or three standards organizations. Is it a job for standards bodies? Yes. Is the industry moving in that direction? Yes. Are they there yet? No.”
Predictably, however, there is more to this than standards. The elephant in the room is the argument for moving from VNFs lifted from physical appliances to cloud-native VNFs, which can be decomposed into microservices and open up the potential for containerisation – virtualisation 2.0. Among other things, this kind of flexibility is the ultimate goal of large-scale initiatives like Three UK’s effort to go fully cloud-native.
4. Inter-carrier federation
It’s not just services specific to a carrier that need to be automated. Services on which carriers collaborate need the same treatment: a latter-day version of NNI.
“There’s plenty of work going on in this area,” says Howard. “Most of it revolves around the MEF’s Lifecycle Services Organisation (LSO). The goal is for multiple operators to define, and operate, or co-operate services. The point is, Telefónica can’t just say to PCCW online, in an automated way, that they need a circuit into seven buildings in Hong Kong. There has to be a service agreement, a business agreement, in place in the first place. So what you’ve got is an effort to federate, or find east-west interfaces, between BSS and OSS systems to enable on-demand services.”
Howard laughs at the suggestion that discussions of NFV security are somewhat muted. “That’s probably an accurate way of putting it,” he says. “It tends to be one of the pieces of functionality that appears next to complex representations of virtualized architecture, alongside “management” or “analytics”.
“But there are definitely people worried about, and working on, the security of virtualised and hybrid physical-virtual networks,” he says. Here, too, he believes there’s a need to move away from traditional methodologies. “In the past, the normal way of building a network function was first you get the basic function working, then you worry about the management, then SLAs, then you worry about efficiency, and then after that you worry: is it secure? The problem with that approach, of course, is the very rapid evolution of new threats. Security isn’t a one-time undertaking any more: it’s a constant.
“You can’t do automation without analytics,” says Howard. “In optical networks, a router network, mobile backhaul network, the mobile core: in every one of those places, you can apply analytics. You machines can learn what normal situations are: by knowing that, they also understand what abnormal is, and the potential remedies. Then you can solve the problem.”
7. The use cases
It’s been said that the virtualisation of carrier networks is a ten to 15 year project. But the first use cases exist: SD-WAN, and, of course, vVoLTE.
Deployments like these are merely the start of a long game. “This is such a huge transformation,” says Howard. “It’s more than moving from TDM to packet. It’s bigger than that, and that was a huge process in terms of equipment and services, all sorts of angles. But this transformation involves a whole change in the organizational structure, involving people’s jobs, and how you design services.”