About a half of a decade after the announcement of the establishment of the Open Networking Foundation, a recent survey of over 3,000 customers internationally by F5 Networks, indicated that just 3 percent of respondents are utilizing SDNs (Software Defined Networks) in production now, while almost 40% have no intentions whatsoever to go ahead with their deployment (37% of the replies suggested that “SDN will become strategically important” within a two to five-year window). Our skepticism over SDN, especially in the public network is well-documented, and a reply to one of our posts from a senior product manager at a large service provider was “the benefits of SDN and NFV (Network Function Virtualization) are more tangible in the data center at the moment.” Although there is no doubt that very large Web 2.0 companies and relatively few other big players have already moved substantially in this direction in order to get needed scalability in clouds, these results from the survey are not exactly that encouraging (no SDN advocates seem to be including this data from a relatively neutral source focused on security in any marketing or PR materials.) The most revealing comments on the subject by F5 regards the matter of additional human resources.
According to the security vendor: “One reason this reluctance persists is because deploying SDN solutions often requires an organization to replace existing infrastructure. This is not only highly disruptive, it’s a financial disincentive, considering that many organizations have significant existing investments in their network architectures.” Therefore, what F5 is pointing to at least indirectly, is that a major impediment to SDN is that new people with different skill sets would have to be added to enterprises.
In general, networking folks take box A and box B, and do not necessarily care how they really work, but they each supposedly provide an x number of functions desired, and they are both plugged together with some kind of standard interface. A major shift needs to occur with SDN, such as the ability of engineers to write Python or perhaps they need familiarity with P4 language. Instead of pieces of hardware, the focus obviously turns to virtual machines, and at a minimum a whole re-education process is required – which tends to violate the universal principal that we have brought up in the past that human beings like to stay with what they know as long as they can – and are fundamentally resistant to any kind of significant change, unless it is absolutely necessary.
While the rhetoric generally regarding an overhaul of the world’s networks to SDN/NFV is still pretty much going strong, people in the industry will continue to disagree on the actual timing of its arrival, not to mention how exactly it would be accomplished. Even when it comes to public landine infrastructure, we have pointed out that major access vendors are moving in this direction. Nevertheless, there is an enormous amount of work that needs to be done to make SDN truly real, if it ever does so, and it must be remembered that a lot of the situations in which a software approach is being considered for FTTP is in totally greenfield situations, which certainly makes an enormous difference in the amount of complexity that is eliminated in not having to deal with an existing installed base of equipment.
[written by Mark Lutkowitz]