& cplSiteName &

NFV Is Down but Not Out

Iain Morris
5/22/2018
100%
0%

In an industry too often guilty of hype and hyperbole, Ibraham Gedeon's bluntness is refreshing. "Our vision is to be able to move these network functions around and make them software-based, and not just take all of that shit and put it on commercial hardware and call it done," says the chief technology officer of Telus, one of Canada's "big three" operators, in fulminating about network functions virtualization (NFV), one of the most hyped telecom technologies of the past decade.

Dreamt up as a concept way back in 2013, NFV is supposed to do exactly what the label says, bringing the virtualization that has already swept through IT into the world of network routers and switches. Essentially, it should mean breaking up the cozy marriage of hardware and software in proprietary network equipment. Software would then take charge of the network. A network function would be a software program you could run over any bit of commodity gear. Costs would fall dramatically. Operators would be able to spin up services with the stroke of a key.

Floored
Can NFV get back on its feet?
Can NFV get back on its feet?

But five years after it was first conceived, a successful divorce of hardware from software has still not happened. Instead, the industry is in the throes of a painful separation that has dragged in all of the family members. Operators still complain they cannot use software from one vendor with equipment from another. Vendors, they add, have ratcheted up software charges to offset hardware losses. Along with the added complexity of NFV, that has eroded cost savings. If NFV is to deliver the benefits it originally promised, operators and vendors will have to reevaluate their entire relationship. (See Virtualization Is Kicking Juniper in the Berries and Cisco Bows to Carrier Demand for Software Outside the Box.)

Gedeon has several gripes about NFV in its current shape. The most obvious is that Telus Corp. (NYSE: TU; Toronto: T) has not been able to realize cost saving targets because of the software fees that vendors charge. "We've dropped the price of hardware to a seventh," he told Light Reading at the Digital Transformation World event in Nice last week. "But if you are adding to the price of software licenses you are not where you need to be." (See Telus CTO: NFV Burden May Cripple Telcos.)

Sitting in Judgment
Ibrahim Gedeon, the chief technology officer of Canada's Telus, has harsh words for NFV vendors.
Ibrahim Gedeon, the chief technology officer of Canada's Telus, has harsh words for NFV vendors.

The cost problem is not just one of licensing fees, however. With NFV have come additional suppliers demanding their pound of operator flesh. "I pay people like Red Hat all of a sudden for OpenStack," grumbles Gedeon, referring to the open source infrastructure platform and one of its chief supporters. Many suppliers also have no "empathy," he says, and are busier trotting out new buzzwords and phrases ("orchestrator of orchestrators" and "artificial intelligence" are two he cites) than tackling basic operator problems. The ability, for example, to spin up a local evolved packet core is still not there, he says.

Far from bringing any kind of automation, NFV seems to have lumbered Telus with additional complexity and operations staff. While overall headcount at Telus rose from 51,300 in 2016 to 53,600 last year, Gedeon attributes the increase to non-telco expansion and says the workforce at the core telecom business has not grown. Ideally, though, he would like to have fewer technology teams. "We've realized a third of the savings. We need to get to a factor of ten," he says. Supplier offerings, it seems, are not helping Telus to realize this target. "Vendors have given us stuff that keeps job security for my 14 teams. How do you make them one team?"

Next page: At an impasse

(33)  | 
Comment  | 
Print  | 
Threaded  |  Newest First  |  Oldest First        ADD A COMMENT
iainmorris
50%
50%
iainmorris,
User Rank: Blogger
5/22/2018 | 6:52:10 AM
Is SDN dead?
NFV might have failed to live up to expectations so far. But is SDN dead?

The question was raised during an conference session at last week's Digital Transformation World event in Nice...
Sterling Perrin
50%
50%
Sterling Perrin,
User Rank: Light Sabre
5/22/2018 | 11:09:15 AM
Re: Is SDN dead?
In a 2017 Heavy Reading survey, 1/3rd of network operators surveyed (globally) reported that they are in some phase of commercial SDN deployment today - so SDN is not dead.

Some still equate SDN with OpenFlow only, and it is likely that OpenFlow is dead in the carrier WAN (I don't know about enterprise). But most of the world moved past the OpenFlow=SDN definition years ago. 

Sterling
iainmorris
50%
50%
iainmorris,
User Rank: Blogger
5/22/2018 | 11:12:47 AM
Re: Is SDN dead?
Thanks, Sterling. I think I recall the survey but will check it out again.

I think one of the issues at Nice was regarding OpenFlow, and more generally about the lack of standards in this area.
Sterling Perrin
50%
50%
Sterling Perrin,
User Rank: Light Sabre
5/22/2018 | 11:20:35 AM
Re: Is SDN dead?
OpenFlow simply did not scale for a carrier network. PacNet was one of the first to deploy SDN in the WAN, and they based it on OpenFlow at the time. (They got bought by Telstra). At BCE, Telstra's Jim Fagan (one of the original PacNet execs) gave an update on the network and he said that they eventually swapped out all of the OpenFlow-based SDN - because of OpenFlow scale/complexity. But it's still an SDN-based network, and they certainly wouldn't say it died when they got rid of OpenFlow!

Sterling
James_B_Crawshaw
50%
50%
James_B_Crawshaw,
User Rank: Blogger
5/22/2018 | 4:06:13 PM
Re: Is SDN dead?
Right. With OpenFlow Pacnet/Telstra managed 500 flows per switch. With their own protocol they can reach 4 million according to Jim Fagan. 
mendyk
50%
50%
mendyk,
User Rank: Light Sabre
5/22/2018 | 11:28:26 AM
Telus more
So Telus's costs are going up because of NFV deployment even as hardware prices drop precipitously. The CTO is blaming suppliers. But isn't the CTO the responsible party here?
iainmorris
50%
50%
iainmorris,
User Rank: Blogger
5/22/2018 | 11:59:45 AM
Re: Telus more
They are making savings, but not as much as the CTO is targeting. 
mendyk
50%
50%
mendyk,
User Rank: Light Sabre
5/22/2018 | 12:09:13 PM
Re: Telus more
The CTO is surprised that vendors would put their revenue/profit targets ahead of their customers' cost-reduction targets (over which the vendors have no say). Seems a bit naive, no?
iainmorris
50%
50%
iainmorris,
User Rank: Blogger
5/22/2018 | 12:17:30 PM
Re: Telus more
It could be seen that way, I guess. Certainly not surprising the vendors want to protect their profit margins, as you say.
Kevin Mitchell
50%
50%
Kevin Mitchell,
User Rank: Light Sabre
5/22/2018 | 11:46:46 AM
Cloud Source Virtualization
NFV continues to confound for broad infrastructure replacement initiatives. A 2016 Heavy Reading survey (sponsored by Alianza) showed that XaaS is also a way forward for CSPs to "augment or replace" network infrastructure. So service providers can cloud source next-gen networks, not just build them. 

See our blog posts referencing this survey:

Heavy Reading: Cloud Defines a New Voice Strategy

Reports: Cloud Sourcing CSP Infrastructure on the Rise


 
Kevin Mitchell
50%
50%
Kevin Mitchell,
User Rank: Light Sabre
5/22/2018 | 4:43:08 PM
SaaS = good partner model
Success-based SaaS model or XaaS model aligns better with being a partner to a CSP. Traditional infrastructure vendors have revenue to protect. How motivated can they be to be disruptive and transformative when it comes to cost? Marginal cost reductions are quite possible when it comes to CAPEX, but will they provide break-through economics? We doubt it
brooks7
100%
0%
brooks7,
User Rank: Light Sabre
5/22/2018 | 5:47:09 PM
Re: SaaS = good partner model
I want to go back to how this succeeded in IT.

The IT companies ignored products from vendors and built their own.  What the telcos want is people to pour Billions of $$ into R&D and pay nothing for it.  If they want integrated platforms that automate, then hire 100 SW engineers, pick available open source, and build shims where required.  That is how the equivalent works in IT land.  When a commercial product is used, it is used off-the-shelf and wrapped in APIs developed by the IT companies.  By trying to build universal standards and get vendors to adapt to them, they have created a fail chain.

seven

 
ASX4000
100%
0%
ASX4000,
User Rank: Lightning
5/23/2018 | 5:07:54 AM
"I pay people like Red Hat all of a sudden for OpenStack,"
No, actually you pay for a service that somebody delivers to you because you aren't able to take care by yourself. If you don't like it, nothing easier than this. Get the people, get the knowledge and do it by yourself. Ohh, and don't forget you need to pay for the skilled people...Umpf! That's the hard way to learn that "opensource" doesn't mean it's for free!
mendyk
50%
50%
mendyk,
User Rank: Light Sabre
5/23/2018 | 9:39:50 AM
Re: "I pay people like Red Hat all of a sudden for OpenStack,"
So some telcos have the same mindset as their customers who want all kinds of services for little or no actual money. Kind of ironic.
yarn
50%
50%
yarn,
User Rank: Light Sabre
5/23/2018 | 10:57:42 AM
Re: "I pay people like Red Hat all of a sudden for OpenStack,"
Perhaps some of the NFV "business case" assumptions were wrong to begin with because network equipment vendors are typically subsidizing the OS software cost through the hardware sales. When hardware and software are sold as unbundled products this is no longer possible, and the cost balance changes completely. Even then it's tough for equipment vendors to support a "white box" or NFV business model because often there aren't any R&D savings that can be passed on since most added value (and revenues) still comes from selling integrated systems. But equipment vendors do apply NFV where there are good application fits for an x86 architecture from a cost/performance point of view.

Recently vendors are starting to offer infrastructure APIs that allow operators to run their own Network OS and directly access the RIB, FIB and MPLS tables, so that might be the way to go. It still remains to be seen how widely applicable, reliable and cost effective that will be in practice for tmost operators. Commercial routing stacks have gone through years of product hardening in thousands of field deployments and countless multi-vendor interworking scenarios. With every new feature you must run a battery of regression tests in a very sizeable testbed to make sure nothing got broken in the upgrade. These are all costs and capabilities that are hard to replicate and easily overlooked. 

Ultimately there aren't many shortcuts if you value quality and reliability. And should things break down it's nice having a vendor you can hold responsible instead of your IT department.
mendyk
50%
50%
mendyk,
User Rank: Light Sabre
5/23/2018 | 11:12:24 AM
Re: "I pay people like Red Hat all of a sudden for OpenStack,"
Vendors and industry experts oversold the economic benefits of NFV, no doubt. The focus on cost saving obscured the main strengths of virtualization, which involve significant improvements in agility, service support, time to market, etc. In other words, business expansion rather than cost reduction. The idea of disintermediating technology suppliers hasn't made any sense for the reasons you note. The decision-makers at the telcos can blame outside forces all they want, but ultimately they are responsible for their company's NFV strategy.
Gabriel Brown
50%
50%
Gabriel Brown,
User Rank: Light Sabre
5/23/2018 | 12:14:17 PM
Re: "I pay people like Red Hat all of a sudden for OpenStack,"
You could argue that in some areas operators have already done well out of NFV by driving down hardware costs on appliances. It's not the operational transformation many want, but is a lot cheaper to buy, say, a classic EPC then it was before NFV. There are many reasons for price movements, but pressure and desire for software-only has certainly played a part.

Time is probably the key metric for NFV/Cloud. Refreshes and new deployments are opportunities and it looks like that is how NFV is making progress... And it is evolving all the time (cloud native, etc.)

The good thing is there is, for the most part, a competitive market. Operators can buy from vendors (including support), they can develop their own, or a mixture of these.
mendyk
50%
50%
mendyk,
User Rank: Light Sabre
5/23/2018 | 1:20:01 PM
Re: "I pay people like Red Hat all of a sudden for OpenStack,"
Right -- which is what makes the whining from some (many? most?) operators so hollow.
Gabriel Brown
50%
50%
Gabriel Brown,
User Rank: Light Sabre
5/24/2018 | 6:39:57 AM
Re: "I pay people like Red Hat all of a sudden for OpenStack,"
Operators that have invested and are operating NFV Cloud (and SDN) seem pretty positive.

They are much less negative than those operators that want pre-packaged vendor solutions for less money. I wonder why.
yarn
50%
50%
yarn,
User Rank: Light Sabre
5/23/2018 | 3:20:34 PM
Re: "I pay people like Red Hat all of a sudden for OpenStack,"
Cheaper to buy is often more expensive to operate, but how many operators actually bother to calculate or project cost of ownership of network purchases over the planned deployment life? I wouldn't be suprised if the CapEx savings on the initial purchase are just a rounding error compared to TCO, yet it is often the decisive factor. But what would the TCO savings be of network equipment that is for instance 20% cheaper but also yields 10% lower network utilization due to reduced port buffering capabilities? The challenge of selling and buying network equipment is how to make these cost/performance differences more transparent. 
Gabriel Brown
50%
50%
Gabriel Brown,
User Rank: Light Sabre
5/24/2018 | 6:47:03 AM
Re: "I pay people like Red Hat all of a sudden for OpenStack,"
Operations and TCO are very important in vendor selection, contract terms, and so on. No doubt operators are often more than a little over-optimistic at the start of, say, a 5-year contract. But it is very much is part of the equation.

The elephant in the room for NFV is public cloud. We kind of have a line-of-sight to a point where many/some network functions will migrate to public cloud infrastcrture. From there you move pretty quickly (logically speaking) to operators paying for network-as-a-service. Maybe a bit speculative and long-range, but to be dismissed at your peril.
Joe Stanganelli
50%
50%
Joe Stanganelli,
User Rank: Light Sabre
5/29/2018 | 11:36:42 PM
Re: "I pay people like Red Hat all of a sudden for OpenStack,"
@yarn: TCO isn't everything, even if it should be more emphasized. Investors in public companies want to see increased profits quarter over quarter -- not increased expenditures. And the executives and lower-level employees making purchasing recommendations and decisions may not expect to be employed at the same company by the time cost savings is realized.
Gabriel Brown
50%
50%
Gabriel Brown,
User Rank: Light Sabre
5/23/2018 | 12:04:43 PM
Re: "I pay people like Red Hat all of a sudden for OpenStack,"
Ultimately there aren't many shortcuts if you value quality and reliability.

Good post yarn

And this is reality for most (all?) operators

And should things break down it's nice having a vendor you can hold responsible instead of your IT department.
rocket101
50%
50%
rocket101,
User Rank: Light Sabre
5/24/2018 | 8:29:37 PM
So does Gedeon from TELUS expects NFV to be FREE????
>>Gedeon has several gripes about NFV in its current shape. The most obvious is that Telus >>Corp. (NYSE: TU; Toronto: T) has not been able to realize cost saving targets >>because of the software fees that vendors charge.

 

You get what you pay for. Even WATER is NOT FREE. Vendors need to make some money, be it NFV or proprietory hardware. NFV Vendors are NOT here for doing CHARITY. If Gedeon expected moving to NFV was FREE, then he is biggest fool of all time. Did he expect vendors to sell Telus NFV for pennies? NFV is not meant to be free, yeah it is cheaper than proprietory hardware, but NOT free. DO NOT EXPECT 99% DISCOUNT, OK?

 Either STAY PUT where you are using the existing hardware or buy NFV and PAY for services lol. 

 

"I pay people like Red Hat all of a sudden for OpenStack," grumbles Gedeon, referring to the >>open source infrastructure platform and one of its chief supporters.

You don't need to go to Red Hat, you can have vanilla Openstack for FREE, but NO support, no hardening etc. Red Hat is here to do business just like any other company. not to do charity. Does TELUS has a mobile plan for FREE???

 

 
Gabriel Brown
50%
50%
Gabriel Brown,
User Rank: Light Sabre
5/25/2018 | 3:49:51 AM
Re: So does Gedeon from TELUS expects NFV to be FREE????
Some of these comments look stark in text in a news report, but it's hard to judge context and tone from here.

Ibrahim Gedeon is a clued-in CTO. I suspect his comments were partly building a picture and reflecting on thought proceses everybody in the industry has had. He isn't naive that is for sure.

He's a good speaker. It's refreshing to have an operator exec articulate these issues in the open.

It may also reflect the frustrations of smaller operators who are more beholden to vendors and who can't (economically) run their own software development programs for what should be "commodity" functions.
mendyk
50%
50%
mendyk,
User Rank: Light Sabre
5/25/2018 | 9:48:06 AM
Re: So does Gedeon from TELUS expects NFV to be FREE????
A follow-up story to clarify some of the points made in the original post would be a good idea. At least it would give Mr. Gedeon the opportunity to clarify his reported comments, which do come off as something short of clued-in.
Joe Stanganelli
50%
50%
Joe Stanganelli,
User Rank: Light Sabre
5/29/2018 | 11:30:46 PM
Re: So does Gedeon from TELUS expects NFV to be FREE????
Even executives can fall for marketing -- and the open-source community's long-time messaging has been "It's free!" The unspoken part of that: "As long as you don't want support lol."
Joe Stanganelli
50%
50%
Joe Stanganelli,
User Rank: Light Sabre
5/29/2018 | 11:28:36 PM
Local EPCs
Yeah, vEPC really fizzled out from all the hype a couple of years ago, huh?

Of course, we can't say we weren't warned for all that hype for those willing and able to read between the lines.  As Gabriel noted in a 2016 Heavy Reading report, revenue from vEPC had a 70-30 split between software licenses and hardware investment -- in favor of software.
rocket101
50%
50%
rocket101,
User Rank: Light Sabre
6/4/2018 | 5:03:27 PM
We've realized a third of the savings. We need to get to a factor of ten
>>We've realized a third of the savings. We need to get to a factor of ten.

Does anyone on this board can share any example where a company was able to realize savings by a factor of ten in ANY business??? If the CTO of TELUS is expecting savings by a factor of ten by moving to NFV, then I feel there is a very basic issue here in the way his understanding is about what is needed to implement NFV OR my undesrtanding is incorrect 100%.

Even a household will not be able to achieve savings by a factor of ten unless folks in the family just drink water for food and eat only salad and that too homegrown most of the time.

How in the world someone and that too a CTO of a company is expecting savings by a factor or ten is beyond my imagination. He is getting 30% savings alreday and that is a huge number by itself. Does Telus or any company offer a phone plan that costs 1/10th or even 1/5th of the going rate just because the network is more NFV focussed?

Please share your thoughts. I just don't seem to come to terms that someone could expect to run an existing business with 90% discount in cost of running the business.

I would like to hear of definite examples if there are any.

 

 

 
brooks7
50%
50%
brooks7,
User Rank: Light Sabre
6/5/2018 | 11:39:01 AM
Re: We've realized a third of the savings. We need to get to a factor of ten
rocket,

 

I think that you are missing the notion a bit.  He is talking about 30% savings over doing the service without NFV.  For a telco, that is a bare minimum for adopting a new vendor.

The big problem will be on the OPEX side, whereas it is pointed out more people are required during the transition.  This transition is the 100% elimination of the old network.  Until then, two staff positions are required. So, even if there are CAPEX savings there is an extreme downside on the OPEX side.

For your example, it would be the adoption of IP PBXes and SIP Trunking.  Not only did one reduce the capital requirements, but you could fire the guy who knows about T-1.  That was a 100% reduction in T-1 guy Opex.

seven

 
Raanan10
50%
50%
Raanan10,
User Rank: Light Beer
6/14/2018 | 1:47:25 AM
‘NFV is not down yet’: save it by doing it right
While obviously not all of NFV's promise has materialized, concluding that NFV has failed to deliver is far reaching. 
 
The values of NFV for service providers and enterprises can still be achieved when adopting the right product and right deployment strategy. D-NFV and uCPE solutions allow quick and efficient rollout of NFV managed services with minimal initial investment supporting an "invest as you grow" model.
 
Special importance should be given to choosing the right uCPE solution. The common tendency to select a solution, using the same criteria employed while selecting centralized NFV, or "jumping on the easy choice" of your current appliance/network-equipment vendor, might not yield the required results.
Atlantis-dude
50%
50%
Atlantis-dude,
User Rank: Light Sabre
8/3/2018 | 1:23:03 PM
depends on what kind of technology team is
built inhouse. If that handles development and support of nfunctions then 10x possible. bUt building such a team is tough and xpensive for telco.
jylliannesiga
50%
50%
jylliannesiga,
User Rank: Light Beer
9/25/2018 | 4:46:07 AM
So does Gedeon from TELUS expects NFV to be FREE????
Some of these comments look stark in text in a news report, but it's hard to judge context and tone from here.
Featured Video
Flash Poll
Upcoming Live Events
September 17-19, 2019, Dallas, Texas
October 1-2, 2019, New Orleans, Louisiana
October 10, 2019, New York, New York
October 22, 2019, Los Angeles, CA
November 5, 2019, London, England
November 7, 2019, London, UK
December 3, 2019, New York, New York
December 3-5, 2019, Vienna, Austria
March 16-18, 2020, Embassy Suites, Denver, Colorado
May 18-20, 2020, Irving Convention Center, Dallas, TX
All Upcoming Live Events