VNFs (Virtual Network Functions)

The Virtual Business Process: A Dilemma

The entire telecom industry is coming to terms with the reality that existing business models are changing dramatically in the virtualization era, but there are strong indications lately that this process is proving problematic for network operators and their vendors alike.

For example, John Isch, director of the Network and Voice practice for Orange Business Services in North America, mentioned in a radio show with our sister site Telco Transformation you can hear in its entirety here, that one of the challenges to the Orange network-as-a-service initiative is getting vendors to accept an on-demand pricing scheme for software licenses of the virtual network functions (VNFs) it delivers to customers.

Orange rolled out its first network-as-a-service offering in the form of a Fortinet Inc. firewall that can be turned up and down via a web portal, once a zero-touch CPE is shipped and installed at the customer premises. Other VNFs can be installed on that box going forward.

But the critical piece is how the VNF software is being billed, Isch explained.

"In this new environment, I don't want the VNF provider to start charging me -- Orange -- for the use of that VNF until a customer turns it up," he said. "When the customer pushes the button, that's when the VNF provider starts charging us and we start charging the customer. If the customer turns it off, all that stops."

That means integration with the Orange BSS, something it has done for its first NaaS offering, but it also means coordination with the vendor and agreement on how the VNF is billed. Finding VNF vendors willing to work in that manner has been a challenge, Isch said.

The 'L' word
Earlier in the month, BT Chief Architect Neil McRae has used his speaker status at Light Reading's OSS in the Era of SDN & NFV event in London to specifically call out issues around licensing, including managing and maintaining large blocks of licenses, worrying about license key expirations, and sorting billing issues. McRae is among many service provider executives chiding the industry's vendors for seemingly replacing the revenues from hardware sales with corresponding software revenue. (See BT: Virtualization Must Include Legacy.)

Travis Ewert, senior vice president of Global Network Software Development at Level 3 Communications Inc. (NYSE: LVLT), admitted in an interview with Light Reading that it's "a little bit wild West right now" where VNF management is concerned, as network operators and vendors alike try to sort out pricing models for licenses, but also how to handle software updates and develop a standard VNF onboarding approach.

"The pure-play software guys use models that are essentially leases -- we don't own the software, we have a right to use it," Ewert says. That means software isn't accounted for in capex, "though there is a crossover point where the cost of a lease is compelling enough to include it in capex."

Want to know more about managing and orchestrating virtualized networks? Check out our orchestration section in the Light Reading NFV channel.

Tying software licensing models to on-demand or usage-sensitive pricing requires a different financial analysis, he notes, and that's a complication for both parties. But there are also issues around lifecycle management of the software, particularly as update cycles get shorter. Telecom's traditional FCAPS model -- the standard approach that includes fault-management, configuration, accounting, performance, and security -- doesn't fit neatly into the VNF realm.

"If you look at those things that are FCAPS in nature, what do you need by way of hooks, day two assurance and more?" Ewert asks. "There are always going to be native capabilities from those VNFs to support that and we might have a VNF manager on top of that. There's a whole certification process us carriers have had to go through historically for network elements that includes things like alarm configuration and performance -- does that change with a VNF? Does it come from the VNF natively or does it sit on top in the management layer? Then we have to have certification with VNF management -- that's all happening under the hood."

Get on board
Those are reasons why a standard VNF on-boarding process was a hot topic at the London OSS event earlier this month with multiple operators and at MEF16 a week later in Baltimore. Standardizing VNF on-boarding was one of three goals AT&T Inc. (NYSE: T)'s Josh Goodell mentioned in his MEF16 keynote address. And it was a significant part of why David Amzallag, head of network virtualization and SDN/NFV of Vodafone, called on telecom vendors to move their software to a cloud native state, from the current cloud-ready position many of them occupy. (See AT&T: MEF Could Catalyze Key Specs and Vodafone: Desperately Seeking Cloud-Centric Tech.)

But as Antonio Elizondo, senior technology expert at Telefónica , noted in a panel at the OSS event, telecom operators are just starting to think about what they need from VNF vendors in the way of information that feeds into systems that provide service assurance, meaning there is still a lot of uncertainty on the network operator side.

"We want to have a model-driven operation," he said. "But that is easy to say, not easy to do. Every VNF has a different requirement." Operators need to have a clear picture of how this comes together to share with vendors -- and their customers, as well, who have their own ideas.

Coming next week: The vendor view

— Carol Wilson, Editor-at-Large, Light Reading

Magnets-are-cool 11/21/2016 | 7:29:05 PM
Re: Get Agile - including with version interoperability The telco model is no single point of failure. Which is the same as the OTT model.

It's the failover time to move to the second/third host/rack/room/datacenter that the OTTs have mastered. In software. And with robust processes and practices.

That's a simplified view of the world, but I think the fact that OTT people control a lot more of their software and hardware stack means they can tune for tradeoffs in ways that telco people can't.


PS: Expectations can be changed. Telco has been providing gold plated service for years...
brooks7 11/21/2016 | 11:23:44 AM
Re: Get Agile - including with version interoperability I would argue that this is the big conundrum for Telcos/MSOs/Wireless trying to become a OTT player.

The nature of OTT is fast fail as you say.  When I was running my little email service (3.5M users), we once did 4 updates in one day (not including the constant security updates).  This was done to fix challenges introduced by customer configurations that we couldn't test in our QA process.

This is how the OTT services work, get it close and put it in front of users.  Fix any issues quickly or back out the changes.

Much more complicated for an infrastructure service.  VNFs as imagined work as infrastucture and their failure might have much more widespread consequences.  People expect their networks to be up 24/7/365 and not to have intermittent performance issues.  OTT players expect people to hit the reload button.  

This has led (and led me to in the email service) to having a much more defined and structured process than a normal OTT player would have.  This included making sure that our update windows for our larger customers were okay with the customer.  Can you imagine Facebook asking when it can do an update?

The situattions are different but telcos can learn from the OTT players.  How much structure?  Less than they are used to but more than 10 guys in an off SOMA who run an update 2x a day.


Magnets-are-cool 11/21/2016 | 12:58:29 AM
Get Agile - including with version interoperability As the OTT players are heard to say, fail fast.

Make changes in small parts and make sure they work. Then rollout it out.

As an industry, we've been stymied by monolithic version for a long time. Time to get agile. Lose some of that 5x 9's mindset. Lose some revenue. But run a whole lot faster.
microcaptechinvestor 11/20/2016 | 8:56:43 PM
Pay-as-you-go // AMDOCS - RADCOM This whole idea of pay-as-you-go is very interesting, and a key reason why the AMDOCS/RADCOM relationship is so interesting especially being tied into ECOMP. AMDOCS has billing expertise, RADCOM service assurance expertise. If CSPs go with these vendors, they can assure SLAs, usage, etc, and track/collect revenue owed from customers.
brooks7 11/18/2016 | 1:04:14 PM

Sounds like at least one person is getting serious about VNF.  Those are many of the right issues.  The next one will be dealing with updates and controlling the correct version numbers of the approved sets of software.  Just something to look at here...I want to use (cringe) Facebook as an example.  Now the basic problem you run into is testing at scale.  Hard enough in a hardware environment (Try generating enough traffic to actually fill a terabit switching fabric and see how much that costs).  The problem here often is application level interaction, which is really tough to simulate. 


Sign In