Juniper did an SDN virtual event yesterday whose title intrigued me; “The Path to SDN”. Since Juniper is one of the vendors whose SDN strategy hasn’t yet been articulated, I decided to attend the event to see if I could get a clue. In a lot of ways, the event ended up mirroring Juniper’s positioning overall—a mixture of insight and missed opportunity.
The opening was given by an industry analyst who cited survey data to set up the position that the world of virtualization and the cloud was a world that demanded things legacy networks didn’t offer. I disagreed with a lot of the data (analyst surveys are like religions; they never agree) but my own data supports the same conclusion so there’s no real foul here. Juniper’s view of what an SDN had to include was also consistent with my own view of reality; they show a functional progression from application-centric interfaces at the top to hardware at the bottom. You could easily map it to my “Cloudifier”, “SDN Central” and “Topologizer” layers. The talk even put OpenFlow in the correct context, which is a means of communicating between two of these layers but far from a complete statement of SDN functionality.
So this is the part where a complete statement could—nay, should—have been made. We can map legacy networking to the same layers as Juniper used for SDN, and then show how each layer evolves, right? That’s not how it went. Juniper focused on the assertion that while virtual resources in cloud computing generally meant multi-tenancy, virtual resources in the network meant AGGREGATE resources. Virtualization that makes a bunch of network boxes look like a single box is the path to SDN. It’s here that I have to part ways with the sense of the talk, because a statement that different from mainstream thought on virtualization demands a justification, and Juniper didn’t provide it.
In fairness, you can make a case for that statement if you have defined a final SDN state, a general strategy to evolve to it, and then some of the specific issues. You can, for example, say that if virtual multi-tenant networks are the future, then anything that abstracts network operation away from device specificity is going to facilitate mapping the network to multiple virtual tenancies. Scale simplification alone would make that true, and I could throw out a couple other good arguments besides. But it’s not my job to make them; this is Juniper’s strategy and they needed (and still need) to make sense of it in the real world. It could have been done, which is why it’s frustrating that it wasn’t.
Juniper is one of the most innovative companies in terms of hardware design. Their gear has a good, solid, reputation. Their positioning has always been muddy and dull, and I think the primary reason is that as soon as somebody draws a chart that has a switch or router in it, the presentation is sucked into being about a box. All of the value of the network universe somehow has to become the property set of this box and thus the story becomes the Story of the Box and not, in this case, the Story of SDN. Shed the box-centricity and this could have been one of the better SDN pitches.
There was an online poll taken at the end, one that offered attendees four optional views representing their take-aways. They ranged from “I think SDNs have promise but I don’t understand them” to “I think Juniper has a good grasp of SDN.” Inexplicably, the results of the poll were shown to all, and the Juniper-favorable choice was the least picked. That’s sad, but it illustrates that if you’re going to do a public event you have to understand the public’s views and needs…and not just the Story of the Box.
There is yet to be a major network equipment vendor with a strong believable story of SDN evolution, and that’s both an opportunity and a problem. Competitors who need such a story, including in particular Alcatel-Lucent and Ericsson, still have a clean slate in positioning themselves and thus have an opportunity to tell the real story. Juniper can still do that too, but every time you come to the plate and strike out it makes it harder to be confident on the next at-bat. And, with buyers in their fall technology planning cycle, we’re rather late in the game now; not too many at-bats left for anyone.