SDN architectures

3 Barriers to SDN Adoption

While SDN makes sense on technological merits, it faces major barriers to adoption. And those barriers aren't technical.

One major barrier to SDN adoption is cultural. Carrier network engineers are currently hardware operators. They deal with dedicated, physical boxes that need to be transported, hooked up, configured with a command-line interface and occasionally visited in situ. SDN will turn carrier networking upside-down; engineers will become software developers, configuring networks using graphical software and writing code in interactive development environments. The only time they'll get up on their feet is if they have treadmill desks.

The job of being a network operator will change. It requires a lot of retraining and redrawing the lines, circles, and boxes on the org chart. (See The Three Faces of SDN and SDN Faces a Human Hurdle .)

Which leads to the second barrier: Power politics. Many network engineers will be simply unable to make the transition to software networking, and they'll be out of work. People resist being put out of jobs. Even tougher for SDN advocates: Some of these people will be upper management who've built their careers around hardware-defined networks. Many of these people are going to see the shift to software networks as a threat to their position within the company. They see the transition to software networks as a death-struggle, and they'll fight against it with every iota of their being.

Sure, hardware-defined networking has its issues. But people whose job it is to solve problems have a strong vested interest in preserving those problems. (I wish I could remember who said that -- it's brilliant.) If your job is to clean the hair out of the shower drain, you're going to fight like a Klingon against any attempt to eliminate shower-drain-hair as a problem.

But what of the benefits? Software-defined networks are more easily maintained, less expensive in capex and opex (allegedly), and more flexible, permitting the deployment of new services faster to customers.

That all appears to be true, but it's difficult to prove -- which is the third barrier to SDN adoption. SDN exists down deep at the bottom of the network, while financial benefits become obvious high up in the application layers. SDN requires foundational network changes whose benefits are indirect and difficult to quantify immediately.

So what's an SDN advocate to do? Start small, deploying SDN in new services. Do fast projects with quick, demonstrable financial return. Enlist allies in the organization where you can find them, and do your best to navigate around opponents without confronting them.

NFV, often mentioned in the same breath as SDN, represents a way to get SDN in the back door of network operators. Virtualizing a network application like a load-balancer or firewall, particularly if it's customer premises equipment, can be a small, local project with immediate financial benefit. Do enough of those and somebody's going to see that virtualizing the underlying network architecture is a great idea too.

As to the managers and senior VPs who fight SDN as a threat to their survival: Some will come around, some will be proven wrong and sidelined, and for others, you may just have to wait for them to retire. Putting arsenic in their coffee, tempting though it may be, will get you the wrong kind of attention.

— Mitch Wagner, Circle me on Google+ Follow me on TwitterVisit my LinkedIn profileFollow me on Facebook, West Coast Bureau Chief, Light Reading. Got a tip about SDN or NFV? Send it to [email protected]

SDNAT 7/10/2014 | 1:45:27 PM
SDN adoption barriers As the article says, Yes, SDN make sense, it is inevitable in the future, and certainly brings lot of benefits. 

In my opinion, some of the key barriers are

1. Risk apetite for new technology:  SDN in general perceived to be for Data Centers. DC is where many of key enterprise apps are hosted. With SDN, multiple changes will be made for DC architecture from technology as well as from overall org perspective. All put together enterprises' perceived risk seems higher. 

2. Fear of unknown: Lack of knowledge about SDN among network engineers/operators/managers etc keeps them away from adopting sooner than later. 

3. Lack of standard: Since different vendors are offering different types of solutions for network agility but are all calling them as SDN the customers are confused and not sure which way the technology is heading. So, enterprises want to wait & watch mode.

4. Currently invested (paid off) network solution is working: Currently deployed network architecture and solution are working fine for the most part and business is functioning. So, no immediate urgency for adopting SDN solution. 


rvrambo 7/8/2014 | 1:00:54 PM
Re: Hyrbid To be frank much better than the original article.  We need absolute and concrete proof of.

1. OPEX savings (  not slides  )

2. Scalability of the solutions to large carrier networks.


SDN adoption is by natute disruptive to the extent that it can effect the company earnings. So any higher ups of the publicly traded carrier networks would think multiple times before signing up on the dottted lines to approve.

Having said that, SDN adoption has to start at a small scale , may at the edge and prove itself  before it can get to core and/or complex services.


If you tell some one that, you can exactly the SAME thing using different set of tools with a great promise of the CAPEX but no OPEX to show for in the next 12-24 months.. that's a bound to fail or rejected at the highest levels.


For example, they said the same kind of things about reconfigurable optical transport few years back and we all know what happened.

Mitch Wagner 7/3/2014 | 12:37:23 PM
Re: Hyrbid Good points. Resistance to SDN isn't foolish, coming from people who demand proof it's ready for prime time before deploying it.
futurephil 7/2/2014 | 7:38:57 PM
Re: and now some specifics... The word "here" is linked. That doesn't show up on this board, though.
VictorRBlake 7/2/2014 | 5:03:20 PM
Re: Hyrbid self editing here, I meant "serious problems" Gotta hate those autocorrect things...
sam masud 7/2/2014 | 4:01:59 PM
Re: Hyrbid If a well understood and popular technology--Ethernet--took a fair amount of time to develop into a carrier-grade service, then we can expect SDN, which is far more complex, will take much longer to be established in carrier networks. But it will be deployed because its capex advantages are hard to ignore.
sam masud 7/2/2014 | 3:58:06 PM
Re: and now some specifics... Did you intend to provide a link in your post?
futurephil 7/2/2014 | 2:42:16 PM
and now some specifics... To sort of pick up where this post leaves off, here are a couple of technical barriers that telcos really are talking about and some advice on overcoming them.
victorblake 7/2/2014 | 10:49:29 AM
Hyrbid First -- I have to say that while you make great points about the challenges to adopting SDN, you do not point out all of the shortcomings of SDN and you exaggerate some of the "claims" or benefits of SDN. No matter how much the controllers do people are still going to have to put fiber or copper in the ground or up on poles. I'm pretty sure that won't be done by software. So I think you've neglected about 1/2 of the challenge of the telecom industry and by capex probably something like 80% to 90% of the capital cost of telecoms. Changing the routers, switches, and OSS to even 100% of SDN -- is perhaps like changing the transmission on a car from a mechanical transmission to a software controlled transmission (much as we did with electrical power control on solar race cars I used to work on). You do not eliminte the need for tires, a frame, and the rest of the "hardware" that makes for a car.

Second, like most technologies I think we are likely to see a hybrid approach where SDN will evolve by at first offering APIs into existing forwarding methods (like MPLS and VPLS VPNs) -- and then latter a variety of hyrbird approaches.

Third, even in the long run, SDN has series problems tha have not been addressed. Not the least of these is that centralized control is very very vulnerable -- in fact it is the exact opposite of IP's distributed architecture and would be subject to the same kind of vulnerability of any other centralized control scheme (like SS7). There's always the swing of the pendulum, but to think it will swing fully to centralized and stay there is to ignore all of telecom history.

While you make many great points, your article implies that senior telecom folks are resisting SDN because it is new or software. I don't think that's the case at all. They are resisting it because it has gaping holes, inconsistencies, immaturity, and -- not least of all because if they tried to build a 100% SDN network today -- they could not -- the products and technology just are not there yet. Despite what you think there are many very senior telecom execs with a strong background in CS -- who have taken the industry to where we are with advanced automation. The number of people required per (any measure such as Tbs) has been on a continuous decline. That's due to the efforts of the people you are claiming don't know about software.
Sign In