Are we in the midst of an SDN hype cycle? History strongly suggests that we are, given the number of technologies that were billed as "The Next Big Thing" in the past, only to fade into irrelevancy within a few short years. Even those that have found a place in enterprise infrastructure – virtualization, for one – have blended into existing architectures rather than taking them over.
By now, just about everyone has heard how SDN will usher in an entirely new era of data interaction, in which optimized application instances will be spun up in a matter of minutes and then just as easily wisped back into the ether so resources can return to the availability pool. It sounds like a wonderful way to run IT, but is it realistic? And even if it is, does the enterprise really want this kind of a change to existing infrastructure?
Some feel that there's not enough hype when it comes to SDN. Ciena CTO Stephen Alexander told the Interop NYC conference this week that through advanced automation and orchestration systems, like the company’s OPn (Optical Packet to the n scale) platform, SDN turns networking from a performance barrier to a performance enabler, one that will ultimately drive user productivity to new heights. The best way to accomplish this, he says, is with a fully open environment that allows users to craft their own data universes using hyperscale infrastructure that can draw from a global resource pool –so long as the user has the means to pay for it, presumably.
If this is the plan, though, it has yet to sink into the brains of those who should be most jazzed about it: network managers. In a recent survey by GigaOm, nearly half of respondents say they have trouble understanding SDN and how it would improve operations within their own networks. I know, that sentence could just have easily been rewritten to say "more than half of networking pros already know how they want to utilize SDN," but the fact remains that with so few deployments underway, we can't know just yet how it will be received by those responsible for overseeing it.
We must also keep in mind that not everyone will implement SDN in the same way. As CIMI Corp. President Tom Nolle pointed out recently, the two SDN deployment models already in place – "shallow" and "deep" SDN, depending on whether you want a simple overlay or device-level implementation using standard protocols like OpenFlow – have been joined by a third, the so-called "long-tailed" model. In this scenario, the enterprise adopts core SDN principles into its policy management stack, where they can be pushed across campuses, out to branch offices, or beyond. The ultimate utility of these and likely future models depends on whether the IT industry can standardize the technology without imposing top-down restrictions on either deployment or operational capabilities.
But again, just because a data environment of this type is possible, does that mean it is desirable? IT Connection’s Mike Fratto wonders if we truly need a five-minute virtual SharePoint instance that can be implemented with just a few mouse clicks. There are valid reasons to manage network deployments carefully, and it seems that most managers would be thrilled just to have reliable automation, better monitoring and troubleshooting, and deeper integration between hardware/software platforms and the processes and products they support.
Of course, if the IT industry is starting to take a rational look at SDN and what it can and cannot do, that would be in keeping with the standard hype cycle. Vendor-driven over-enthusiasm always precedes a pessimistic backlash and, finally, acceptance that the world is not coming to an end. And this process usually plays out before the technology has even proven itself in the field.
SDN could very well usher in the fully virtualized, on-demand data center, but it’s up to CIOs to determine if they really want to take it that far.