AT&T: Data Drive Demands SDNs

MUNICH -- Ethernet Europe 2012 -- The data networking world needs to go through the same transition to software-defined networking that the voice world underwent two decades ago, said AT&T Inc. (NYSE: T) optical and Ethernet guru Margaret Chiosi here on Tuesday.

Chiosi compared the current suggested move to OpenFlow, with its more centralized routing intelligence, to the transition voice networks made to SS7, when voice call routing was handled more intelligently through a separate signaling network of signal transfer points and service switching points. (See Why Verizon Is Keen on OpenFlow.)

"By taking the network intelligence out of the router and put it into a server, you can ride the wave of the server market, which is dictated by Moore's Law," Chiosi said. To ride that wave in the router world, she adds, requires installation of new blades, physical swaps and then the software upgrades, she added.

"There's a lot of skepticism," Chiosi admitted. "But the world has done this before -- we did it with the voice network."

Moving to a software-defined network is one of the key transitions the industry needs to leverage the move to cloud services, Chiosi said. It would enable network applications to request and manipulate services provided by the network and allow the network to expose its state back to the applications.

The basic approach to wide-area networking also needs to change, she added. Today's WANs are designed for multi-year contracts, not dynamic, on-demand services. Service providers need to define the types of services needed to support applications by latency and availability and then offer those services on an on-demand basis for set durations of time.

"The physical provisioning still needs to take place, and that takes time, but once you have that, the service provisioning can be instantaneous, and customers should be able to pay as they go," Chiosi said.

— Carol Wilson, Chief Editor, Events, Light Reading

DCITDave 12/5/2012 | 5:36:13 PM
re: AT&T: Data Drive Demands SDNs

Is what's good for a voice network (static, connection-oriented, low bandwidth) always make sense for data networks (dynamic, bursty, high bandwidth)?

Perhaps OpenFlow is the way to go but I'm just not sure the analogy makes sense in this case.

cnwedit 12/5/2012 | 5:36:09 PM
re: AT&T: Data Drive Demands SDNs

In this case, the analogy is based on where the intelligence for the service lies. With the move to SS7, the telecom industry opened the door to a new world of services - things such as Caller ID become possible, along with better routing of calls and more efficient use of network resources.

I think Chiosi is saying this is a similar change with similar potential benefits, even though the services themselves are very different. She's trying to address OpenFlow skeptics with her remarks - I don't know if this analogy will work for them, but I get it.

DCITDave 12/5/2012 | 5:36:07 PM
re: AT&T: Data Drive Demands SDNs

I kind of doubt they'd be altrustic, but I think I get what opto is saying. Service providers still have a mindset of managing scarcity vs. creating abundance. They can make money doing both, but the latter is more lucrative long-term and requires more up-front risk (i.e., build platforms and let others innovate vs. owning every step of the process).

I still think service providers could make more as dumb pipes than as intelligent networks, but I understand the shareholder pressure to produce "growth" and that usually requires them to tack on a fee to something that should be included in basic service.

optodoofus 12/5/2012 | 5:36:07 PM
re: AT&T: Data Drive Demands SDNs

Here's the thing: service providers are only interested in how service providers can make money.  They want to build an intelligent network and charge through the nose for the services (like they did with caller ID).  They are stuck in 1980's thinking, but the world has moved on. The time for an intelligent network has passed, and it ain't coming back.

If you haven't read it in a while, go back and re-read The Rise of the Stupid Network.  It was a fantastic analsysis of the state of telecom, and its predictions were uncannily accurate.  We watched the wireline itelecom ndustry crumble because endpoints became smart and network intelligence was no longer needed.  Here we are 15 years later, and the same thing is happening to the wireless idustry.  Why the delay?  It took that long to build affordable. compact intelligent wireless endpoints (AKA smartphones).  But he end gam eis every bit as inevitable in wireless as it was in wireline.

Given this, I find it hilarious that AT&T is suggesting that we need to re-create something like AIN but for data networks.  Talk about not learning form the past!  I guess when you are a hammer, every problem looks like a nail.


shygye75 12/5/2012 | 5:36:07 PM
re: AT&T: Data Drive Demands SDNs

No doubt the growth imperative leads companies to focus on shorter-term results vs. the long run. That's true pretty much across the board, and across all industries. Network operators are a convenient pinata, but there are plenty of candy-filled entities primed for whacking with a stick.

shygye75 12/5/2012 | 5:36:07 PM
re: AT&T: Data Drive Demands SDNs

"[S]ervice providers are only interested in how service providers can make money." 

What else should they be interested in, for instance?

joanengebretson 12/5/2012 | 5:35:55 PM
re: AT&T: Data Drive Demands SDNs

I think there's definitely a need for what Chiosi is describing. A key benefit of cloud computing is to be able to add computing resources as needed. But if you can't adjust the bandwidth between the cloud and the customer in an equally dynamic manner you've got a problem.

Soupafly 12/5/2012 | 5:35:45 PM
re: AT&T: Data Drive Demands SDNs

The article presents a very simplistic soundbite led approach to the SDN arena. Not having heard the keynote its difficult to respond without overly generalising.

The conversation comes down to a benefits deliverable. De-coupling the control and the data plane has been happening in routers (by and large) for years. The seperation (in part) faciltates some of the "Non Stop" protocol implementation. (NSR, NSF, GR, etc)

Having said that the security implications for such an implementation are simply ghastly! To take control of a conventional server (because thats what chioisi seems to be referring to in her article) is infinitely easier than hacking into a proprietary router OS. My assumption here is that SDN will exploit existing ATCA & other hardware modules and OpenSource languages for its base build.

She has also failed in her article to mention the role of Optical networking. Given that layer 1,2 and 3 are blurring, how does that come into her thinking?

At the moment the key architecture and deployment windows favoured by SDN seem to be DC interconnects. Thats not what chioisi seems to be referring to either...

So as I stated at the start. Many questions, few answers and allot of promises (the chioisi logic) but few if any CEO/CFO would put this tech into a production network. We are years away from this been anything other than a vision.

Sign In