With recent movement the last few months in the SDN controller space, it appears that 2014 might be the year where controllers duke it out for dominance. Certainly we haven’t seen the end of the virtual switching wars of 2013 yet; with a bunch of virtual switches both open-source and commercial in play (Nicira/VMware's NSX with its vswitch component; NEC’s PF1000V for Hyper-V environments, the Cisco NX1000V and that new Application Virtual Switch in Insieme’s announcement; Open vSwitch's new version 2.0; and FlowForwarding.org's LINC softswitch).

But while virtual switches are strategic, a new battle appears to be brewing. My partner, Matt Palmer, talked about what we heard from the community in his recent post on ON.LAB and ONOS. And that prompted me to write a state-of-the-union article summarizing where we are today with controllers, how things might play out in 2014 and beyond, and what enterprises, cloud providers, and service providers must know and should do.

The controller is a strategic control point in an SDN network — it is essentially the brains of the network. And no networking vendor is going to cede control of their network equipment to a third party, because it puts them at risk of being dis-intermediated and subsequently pushed out, or at the very least, marginalized.

And so, there are two possibly stable models in an ecosystem: one where dominant networking vendors provide their own controllers that orchestrate their equipment and possibly others, or another where a “safe, trusted” third party provides an open controller that all vendors (or at least a majority) support. The latter model is of course what the OpenDaylight Project (ODP) is proposing. ODP has gained substantial vendor support over the last six months, with an increasing number of vendors jumping in to contribute their expertise and code. This open controller will also enable end-users like cloud and service providers to build on top of the foundation, adding their proprietary "secret sauce" that allows them to provide differentiated offerings in the marketplace.

To understand the value of open-source controllers like ODP going forward, we need to understand what’s happening in the commercial space. Let’s look at some well-known networking vendors and examine who has a controller today (or is planning to launch one):

Of the above commercial controllers, the ones to watch are Cisco’s and VMware’s, since they each are leaders in networking and virtualization, but I would also be keeping my eye on NEC, HP, IBM, Nuage, and Juniper to see if they are able to expand their spans of control beyond their own equipment. The other controllers are very solution-specific (primarily closed-network virtualization solutions) and unlikely to grow beyond single-vendor ecosystems in the near term. 

On the other end of the spectrum, to balance out the commercial offerings, and which in theory should keep commercial vendors in check, are the open-source controllers. And in the open-source world, there is certainly one granddaddy today — ODP. With a large support base of numerous vendors and a growing independent developer ecosystem, it seems like the one to beat. However, it has been challenged in pulling together disparate code bases and diverse modules from multiple parties. It’s no surprise that integrating such a beast is a challenging task. The Hydrogen release, originally targeted for early this month, has been pushed out and hopefully will be tested and released before the OpenDaylight Summit on Feb 4, 2014.

Newer open-source controllers out there include OpenContrail from Juniper and ONOS from ON.LAB (not released yet), adding to the list of open-source controllers that have been around for some time, including NOX, POX, Beacon, Floodlight, Trema, Ryu etc. Trema is a project within NEC and is still active but has seen limited traction in the commercial world. Ryu is backed by NTT and actively worked on, particularly around its OpenStack integration.

What Next?

In part II of the post, I compare the differences between the three new major open-source controllers and provide my opinion of how this will play out  in 2014. I also discuss what service providers and enterprises ought to consider as they move ahead with their SDN and NFV strategies.

On to Part II!