x
SDN architectures

XO Takes a Shine to OpenFlow

NEW YORK -- Packet-Optical Transport Evolution 2012 -- XO Communications Inc. is intrigued by OpenFlow and is giving all its vendors suggestions about moving towards software-defined networking, CTO Randy Nicklas said in an afternoon keynote here Thursday.

At the same time, Nicklas is realistic about OpenFlow.

"If you're asking me what's the startling new revenue driver that's going to come out of OpenFlow ... I'll say it's not clear there will be one," Nicklas said.

What's so interesting about OpenFlow, then? Nicklas likes its openness, which is the aspect touted by startups and organizations including Big Switch Networks , Nicira Networks Inc. , the Most Open Organization and the Open Networking Foundation .

Having a centralized control plane isn't so unusual; people used to do it with ATM and Frame Relay, Nicklas said. But it's always been done using vendor proprietary software. OpenFlow would be the same idea under a universal protocol, working with any vendor's hardware. And that leads to cost savings, because one type of controller could be applied to any vendors' switches, theoretically.

"It's not enough, but we would say it's a good start," he says.

Beyond OpenFlow, Nicklas would like to see software that can extract some of the information about what's going on inside switches: What the queue depths are, for instance. That kind of information could be fed back into OpenFlow controllers as input for switching decisions.

Nicklas's vision of software-defined networking bears some similarity to Cisco Systems Inc. (Nasdaq: CSCO)'s contention that the real power of software-defined networking lies in unlocking the state information that applications can't identify. (See Cisco Links SDN & Policy and Cisco Broadens Its Software-Defined Networking.)

Nicklas referenced Cisco, in fact. XO is using the ASR 9000 edge router and the Carrier Packet Transport 600, Cisco's packet-optical system, in its metro rollouts -- and he's brought up these software ideas with Cisco. "They're saying it's very interesting and they want to talk more, and that's all we would expect at this stage," he says.

He prefers OpenFlow to the Path Computation Element (PCE), the Internet Engineering Task Force (IETF) standard that's been in development since 2006. PCE involves telling a lot of network elements what to do, whereas OpenFlow is "very simple," he said. "It says: Here's a flow-able entry, and here's the set of messages you can send to an OpenFlow client ... It says nothing about how to compute a path."

— Craig Matsumoto, Managing Editor, Light Reading

Pete Baldwin 12/5/2012 | 5:32:40 PM
re: XO Takes a Shine to OpenFlow

Hi Obaut -- I don't think so, not yet. Even Nicklas made it sound like this is a dreamy next-step that's going to take a while to happen.


Intriguing idea, though, having that feedback loop between the controller and the network -- more automation in the network.

obaut 12/5/2012 | 5:32:40 PM
re: XO Takes a Shine to OpenFlow

"Beyond OpenFlow, Nicklas would like to see software that can extract some of the information about what's going on inside switches: What the queue depths are, for instance. That kind of information could be fed back into OpenFlow controllers as input for switching decisions."


Good point - it's time to move from discussing openflow "framework" to deploying practical features, such as above. On this particular point - is any vendor/SP actually supporting this kind of functionality (in the context of more flexible interworking between NE hardware and switching control functional layers)?

HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE