There is no doubt that the enterprise is looking forward to SDN in a big way. With an abstract network at its command, end-to-end data infrastructure can finally exist on the virtual plane, ushering in a level of flexibility that the industry as a whole is only beginning to comprehend.
But the question is not whether we want SDN, but how to get it. Or, more to the point, how much disruption is the enterprise willing to put up with in order to achieve this networking nirvana?
The answer, it seems, is not much.
Enterprises reluctant to make real moves towards SDN
According to a new survey from Piper Jaffrey, only about 12 percent of CIOs are planning to replace traditional routers and switches in order to implement SDN in the coming year, while more than half said they are not planning for SDN at all. The remaining third say they are evaluating their options. Oddly, however, a large majority identified networking as the facet of the data environment that is most in need of a refresh, which causes one to wonder: If not SDN, than what?
The answer can be found between the lines. Enterprises are eager to deploy SDN. They just don’t want to undergo a full rip-and-replace of network infrastructure to get it. As the Wicked Witch said to Dorothy: “These things have to be done delicately, or you hurt the spell.”
Gradual SDN migration with network overlays, open switches, and open source
It is for this reason that many top vendors are pushing network overlays and open switching formats. The idea is that the enterprise can replace expensive devices with commodity hardware as part of the normal refresh cycle, and then implement SDN software as they see fit. Dell is a prime example, having recently teamed up with Cumulus Networks and Midokura to provide a Linux-based OpenStack distribution within a virtual overlay atop a commodity switch. This is a pretty big step for Dell, considering the Cumulus OS is a rival to Dell’s own network operating system and essentially robs the switch of most of the bells and whistles that have driven substantial revenue streams into company coffers for the past two decades.
Of course, the desire for a go-slow approach to SDN is music to the ears of longtime open source proponents like Red Hat. The company is quickly lining up a wide range of support for its RHEL OpenStack Platform 5, the latest being Big Switch, which recently obtained certification for its Big Cloud Fabric. This will allow the enterprise to introduce SDN slowly and carefully through commodity switches, while at the same time laying the groundwork for advanced OpenStack releases like Neutron, Nova and Juno that enable dynamic network configuration, single-pane management capabilities and support for advanced applications like DBaaS, virtual desktops and Big Data analytics.
The proprietary switch vendors are not blind to this reality either, but there are limits as to how far they are willing to go in support of diverse SDN environments. Brocade, for example, is okay with its Vyatta platform working across a range of physical and virtual networking platforms via its ties to the Open Daylight project, but it is still the basis for an end-to-end SDN platform ostensibly designed for Brocade switching hardware. In this way, Brocade hopes to serve two types of enterprise: those that are looking for a full conversion to SDN, either in legacy or greenfield deployments, and those looking to implement the technology gradually.
In a way, it’s kind of comforting knowing that even in the new virtual architecture, issues like open source vs. proprietary are still with us. No matter how ephemeral the data environment becomes, some things remain the same.
And for those who are sweating the coming transition to virtual networking, relax. The change will not be as rushed or dramatic as it once seemed. The enterprise is still in the driver’s seat when it comes to SDN, and it will happen as quickly or as slowly as budgets and competitive pressures mandate.