Mike Fratto
Summary Bullets:
- Many enterprises are struggling to find a reason to invest in SDN when what they have in place works well enough.
- There will not be a killer app for SDN. Rather, enterprises will adopt SDN as their data centers needs evolve over time.
I recently attended and presented at the SDN and OpenFlow World Congress in Bod Homburg, Germany, and while I was there, I had a chance to talk to some enterprise IT attendees who were investigating SDN for their organizations. The common question through most of the discussion was: “Why should I use a SDN?” Luckily, I started my talk with that very question.
Enterprises need a reason to move to SDN and that reason is not yet evident. Certainly there are early adopters that have a critical need for something different in their data center, but for the majority of enterprises, what they have in place right now is sufficient. The reality is that in many enterprise data centers, physical and virtual assets are relatively fixed and are not moving around hypervisors like a conga line; nor are business line managers demanding applications be deployed yesterday or we will fail. IT is nowhere near the point where they can reliably let business line managers bring up new applications on demand. Enterprise data centers simply are not that dynamic and the integration that exists between hypervisors and networking is good enough.
The latest reason to move to SDN is because it takes network administrators a very long time – days, weeks, months, depending on who you ask – to do something simple like provision a VLAN. That’s a disingenuous claim because: A) those network admins have other things on their plate and aren’t sitting idly by the phone waiting for a change order to come in; and B) experienced IT shops have change control procedures to ensure that changes are made properly and they can be backed out if something goes wrong (because, ya know, Gremlins!). The other end of the spectrum is cowboy IT, where network engineers make changes first and then figure out what went wrong later. In other words, there are reasons why IT takes time to make seemingly simple changes. By the way, the same can be said for adding a new server, application, storage, or any other IT component.
What’s the App?
If I knew what the killer SDN app is I’d be looking for some VC backers (don’t tell my manager). In fact, I don’t think there will be a killer app or feature. I think the migration to SDN in the enterprise will be gradual and trail product and service innovations as they trickle out into enterprise products and as VARs and consultants get experience with SDN technology and can point to customer deployments. Remember, it took a long time for server virtualization to make headway in the data center. From 2001 to ~2006, there was not a good reason to migrate servers to hypervisors, and today virtualization is commonplace, though 100% virtualized data centers are still rare.
SDN is going to follow the same path. As more applications and use cases are available, the reasons for enterprises to delay migrating to SDN will be fewer and fewer. In 20 years, newly minted network administrators will wonder how it was ever any different.
