NFV Strategies

SDN Is Hype & NFV a Faux Pas – Telecom Panel

PARIS -- MPLS, SDN and NFV World Congress 2018 -- There must have been something in the coffee in Paris this week.

Panel sessions at trade shows are often a reality check for the industry, but to hear one deliver such a negative assessment of a technology in the limelight is unusual.

Executives from Orange and Telefónica on the telco side, Nokia Corp. (NYSE: NOK) and RAD Data Communications Ltd. in the vendor community and Google (Nasdaq: GOOG) from the Internet world were asked here in Paris to consider whether SDN and NFV were mainly hype or reality. And their discussion came down firmly on the side of hype.

That is a bit of a wake-up call for the wider industry. SDN and NFV have now been around for several years. Orange (NYSE: FTE) and Telefónica, moreover, are typically seen as pioneers when it comes to investment in these technologies. Telefónica was one of the first European giants to make a big commitment to virtualization when it announced its Unica program in 2014. Orange said it had moved into the "industrial phase" of NFV deployment in October. (See Orange Moves to 'Industrial' NFV Phase, Will Start in Spain and Telefónica Unveils Aggressive NFV Plans.)

But that did not prevent Stephane Litkowski, a network architect with Orange, from questioning the rationale for spending money on SDN and NFV technologies.

"In my view SDN is still hype and not really a new concept," he told an audience at this week's MPLS, SDN and NFV World Congress in Paris. "The centralization of network management is something we've been able to do for years. It seems we are never really investing in new things but reusing concepts because we can never find a good trade-off. Centralization has its advantages but there is also a need for distribution. I would never set up a network when I have only a single controller because what happens if it fails?"

Israeli vendor RAD also joined in the SDN and NFV bashing, with Yaakov Stein, the company's chief technology officer, suggesting SDN is a "bubble" and questioning NFV's cost benefits.

"When you are doing network functions out in the network, the question is whether you are really saving money by using generic instead of optimized hardware," he said.

To compete with the OTT players, telcos need to be nimbler by accelerating network automation. Join us in Austin, Texas from May 14-16 for our fifth annual Big Communications Event as we tackle challenges like automation. The event is free for communications service providers -- secure your seat today!

RAD, of course, has legacy as well as SDN and NFV products in its portfolio, but Litkowski agreed with Stein's remarks. "The cost of the power needed with generic hardware is not really economical," he pointed out. "NFV is a reality but is it the reality we wanted?"

On a more positive note, the Orange executive suggested the use of field programmable gate arrays (or FGPAs), which are customizable integrated circuits, might help to make NFV a more welcome reality. "With network functions we have very special requirements and the main pain point is still performance and the determinism of packet processing," he said.

For Nokia's Wim Henderickx, the main challenge is on the applications side. "The applications people still look at the network as though it is from the past," he said. "They are still relying on Layer 2-based resiliency mechanisms and applying the physical world in a virtual environment. We are missing specifications on how apps should run in the cloud and as a result there is complexity in the onboarding element."

Coming several hours after Google's Vijoy Pandey had talked about the Internet giant's investments in intent-based networking tools, to support further automation of its data centers and backbone networks, this was all rather worrying, despite the jocular tone of the panel discussion. If they are not about cost, SDN and NFV are supposed to aid telco agility and speed to market. Yet technology executives are still voicing doubts more than four years after Telefónica first mentioned Unica. For the industry's sake, one must hope the discussion has moved on this time next year. (See Google Has Intent to Cut Humans Out of Network.)

— Iain Morris, News Editor, Light Reading

<<   <   Page 2 / 2
komatineni 4/16/2018 | 8:09:59 PM
Perspective and expectation IMHO

It all depends on what was the original expectation from SDN/NFV? If it was about changing the world, removing the hunger, solve all human problems, innovate new solutions and reduce capex/opex then it's busted.

If the expectation was to automate certain routine tasks, move over to COTS from a branded heavy ATCA/Custom based architecture & have some capitol efficiencies with flexible models - met half way.

Bottomline is it all depending on the business case and underlying needs/wants/challenges. In APAC, a majority of the TELCO's are not run by engineers but by accountants & Telco's need to adapt to strict regulatory requirements. When WhatsApp fails for an hour, there is no grilling by a gov or regulatory penalties. When a mobile fails for 10 min, the operator is liable for heavy penalties. Now, from risk mgmt perspective what makes more sense? Go with some tested solution or try to have a big 'programmer' team and explore DIY platforms for virtualizations? I dont think the market prefer the latter.
Gabriel Brown 4/16/2018 | 9:32:31 AM
Re: I too was there (in fact, on the stage!) Good comment Yaakov. 

Distributed virtual data-plane is also an in issue in mobile networks. This scales from the enetprise and cell site level to the transport aggregation network.
Y(J)S 4/15/2018 | 1:38:05 AM
I too was there (in fact, on the stage!) In panel discussions messages are distilled into soundbytes, which are then further reduced in published reports. The resulting soup may not capture the nuances of a subject intricate enough to warrant a panel in the first place.

The statement attributed to me about SDN being a "bubble" was not mine. I was quoting Bruce Davie at this same event several years ago. In fact, when the audience applauded I reminded them that the applause was for Bruce, not for me. As reported in previous comments, no-one on the panel thought SDN was "pure hype". although all thought that there was a pendulum going back and forth between distributed and centralized intelligence.

Regarding NFV, my point was that there are centralized software functions that were previously in branded boxes and now run in COTS servers. This increases competition and thus saves CAPEX, but is not a radical change. The radically new idea is the virtualization of non-centralized data plane functions, an idea championed by RAD as DNFV, and now widely implemented as a vCPE. Here the CAPEX often increases due to the whitebox platform not being optimized, and thus the economics depend on whether OPEX savings can offset this increase. Some service providers emphatically say yes, while others are having problems with the arithmetic. In my talk a few hours later I explained how pluggable accelerators can help tip the scales in favor of NFV.

Yaakov Stein

Editor34868 4/13/2018 | 11:12:15 AM
Re: Quite a different message... I was there too. According to my notes, they said it's being implemented but also that there are still teething problems. My main take-away from that session was the issue of power consumption: virtualisation implies implementation on general-purpose computing hardware, which can take an order of magnitude more power than dedicated hardware. (Just think how many transistors you have to switch in order to decode and execute an instruction that changes one bit.) So there's a tension between "virtualisation" to make things more agile and "concretisation" to save power.
yuri_gittik 4/13/2018 | 6:31:48 AM
Quite a different message... I was in the audience of this panel session, and took away a different message, so reading this article was a big surprise for me. Did we attend the same session?

All the panelists clearly agreed that NFV (and SDN) is already a reality, not a hype. The transformation journey has started, and for sure, there are some challenges on the way. RAD has been leading edge NFV from the very beginning (with DNFV approach that was emerged as vCPE/uCPE solution), so we do see this journey from inside. Indeed, there are challenges and they were addressed and discussed during the panel, including possible means to mitigate them (e.g. modular HW acceleration for boosting WB performance).
<<   <   Page 2 / 2
Sign In