• 01/17/2014
    10:51 AM
  • Rating: 
    0 votes
    Vote up!
    Vote down!

Don't Let SDN Infatuation Go To Your Head

The networking industry is all a-flutter, with vendors hastily breeding new software-defined networking platforms and controllers. Smart enterprises will show a healthy fear of commitment.

Just say software-defined networking, and I begin to feel a bit like Elizabeth Barrett Browning, she of the "How do I Iove thee? Let me count the ways" fame. After all, what's not to like about a technology (or is it a framework?) that promises to reduce network capex/opex and likely gives application development and new services a boost? Enterprise customers, in particular, should be especially enamored, since much of SDN's promise is focused on its deployment in the datacenter.

But so diffused and amorphous is the current SDN vendor market that, depending on who's counting, there are at least a couple dozen suppliers out there. That's pretty impressive when you consider that economically things are really just puttering along.

I had hoped that Cisco's unveiling of its Application Centric Infrastructure (ACI) for datacenter networking, when seen in the light of AT&T's Supplier Domain 2.0 announcement, might be an indicator that Cisco had its finger on the pulse of the datacenter market, offering customers a reasonable roadmap for SDN.

That apparently was wishful thinking. In fact, if you buy into a report on AT&T's Domain from financial services firm MKM Partners, it isn't good news for Cisco at all. Of the 10 vendors tracked by MKM Partners, only Cisco received a "negative" in terms of the long-term impact of AT&T's Domain 2.0 technology vision for the datacenter and wide area network. By contrast, Alcatel-Lucent and Juniper, rivals of Cisco in the switch and router market, were given a 50-50 chance to win AT&T Domain 2.0 business. 

Why isn't Cisco ACI in the running for AT&T's future datacenter business? The short answer, according to the report, is that AT&T is looking for a more open solution than Cisco ACI. The report predicts that AT&T isn't expected to deploy SDN in the datacenter before 2015, and it will "likely be at least another year beyond that or 2016 before a WAN SDN controller is deployed." 

One doesn't have to look far and wide to understand why even a huge company like AT&T is being cautious about SDN. Take, for example, the SDN controller, which is the brains of any SDN deployment (or the heart, if you're a Liz Barrett fan). Roy Chua at SDNCentral has published a concise list of SDN controller vendors. It's obvious from the list is that there are whole lot of controllers in various flavors, with many vendors producing more than one (Cisco has triplets).

One would need to be a diehard optimistic to believe much has changed from six months ago when networking expert Jim Metzler agreed with distinguished Yahoo architect Igor Gashinsky's view that "SDN is not ready for production in the typical enterprise network," but also warned that "if IT organizations aren't familiarizing themselves with and testing SDN today, they are going to be left behind."

That sounds like a pretty fair assessment and good advice. Enterprises may not want to place their bets just yet on any of the available SDN choices. When the big technology companies, such as AT&T, Amazon, and Google, make their choices, it will bring a level of discipline to the market. At this time, there is a lot to like about SDN -- except the reality. And that's not true love.

Sam Masud has covered the telecom industry for more than two decades as an analyst and editor. 


SDN -- not if, but when

The future of the data center -- and the network -- is to become a more programmable, software-defined and software-controlled entity. I lean toward the better-be-thinking-about-this school than the skeptics advice that there's no there, there yet. It's not a question of if, but when.

Re: SDN -- not if, but when

Agree that it's only a matter of time, but it seems that right now it's really the deep-pocketed industry players, e.g., Google, who have in-house tech expertise who are deploying SDN. In Google case, it's for interconnecting data center. Also, Alcatel-Lucent's SDN venture, Nuage, just recently announced that the Nuage SDN solution has been selected for a backup network in the University of Pittsburgh's Medical Center. The fact that UPMC's is first going to use SDN for as a backup network tells me it's taking a cautious approach.

Re: SDN -- not if, but when

Agree Charles. SDN offers too many compelling business benefits (costs savings the least of them) to be ignored and Cisco has most to lose. However, as with any new technology and in this case, a true paradigm shift, management is key. SDN/NFV will not gain broad adoption in the enterprise until manageability is addressed. The management ISV community needs to be engaged. Some related commentary here:

Market Adoption
"When the big technology companies, such as AT&T, Amazon, and Google, make their choices, it will bring a level of discipline to the market"

The only thing is, at least two of these three if not all three have the resources and skills to create their own customized solutions rather than neccesarily having to make any choices from the commercial offerings you and I might be considering. If they go "white box", should we follow? Perhaps not, unless we also have their controller code...

Re: Market Adoption

You'll not get any argument from me on that because these hi-tech companies have resources that others don't--including many enterprise customers whose core business might be manufacturing or something else. Of course, the SMBs have even less resources.

The white box "solution" is a related but different issue. I don't know how it will pan out, but here's a crazy, or perhaps not so crazy thought: what if systems integrators/VARs were to get into the business of tailoring white boxes (either servers or switches or both) for the commercial market?