One of the most exciting aspects of working with OpenFlow implementations and migrating to software-defined networking (SDN) is that "the networking community, which has traditionally been controlled and locked up by vendors, is growing stronger. There's a sense of community now that wasn't there before," said Brent Salisbury, lead network engineer at the University of Kentucky. "The innovation enjoyed by the open source software community for decades is coming to networking."
For example, the Open Networking Foundation's board is made up of end-user companies rather than vendors. With OpenDaylight, vendors are trying to build out a community of users and developers. And the Open Networking Summit is also expanding in size every year. These are all helping bring the networking community together to work on challenges.
During its time working with early OpenFlow implementations, Rackspace has also discovered a few noteworthy things.
First, Rackspace hired a different skill set to work with OpenFlow. "We tend to be a lot more focused on systems on our SDN side than network-focused. It was easier to find strong systems people and show them what they needed to know about networking than to go the other way," said Brad McConnell, principal architect at Rackspace.
Second, it's critical to realize that OpenFlow itself doesn't give you a new feature, McConnell pointed out. "At the end of the day, features are what customers pay for. Only adopt OpenFlow because you see how it can give you some significant gain in your data center. It could be that you have a very basic framework that you can use to make a new feature that limits how Neighbor Discovery works on IPv6 … since it doesn't really have a tool to do it on any of the old switches in my data center today. Stay outcome-focused. … Don't just jump on the trend."
Rackspace also just joined the Open Networking Foundation, hoping to hear more about where things are headed and because the infrastructure team is interested in the primary use cases -- especially in the enterprise space. "One of the most common use cases is going to be around traffic engineering, because you have this thing that sits outside your network and computes all of your paths for you and tells your network what to do," McConnell said. "That's definitely a space where having real-time knowledge about the entire network's state is advantageous."