At Martin Casado’s Network Heresy blog yesterday, a guest post was offered by Andrew Lambeth, who once led the vDS distributed switching project at VMware but is now, like Casado, ensconced at Nicira. The post was titled provocatively, “Networking Doesn’t Need a VMWare.”
It was different in substance and tone from Casado’s posts, which typically are balanced, logical, and carefully constructed. I appreciate those qualities. Words matter, and Casado invariably takes the time to choose the right ones and to compose posts that communicate complicated ideas clearly. Even better, he does so without undue vendor bias.
Maybe he’s really a shrewd master of manipulation, but I always get the impression Casado is sincere, that he means what he says and says what he means. One actually learns something from reading his blog. That’s always refreshing, in this industry or any other.
Defining (or Redefining) Network Virtualization
As I said, the post from Lambeth was a departure in more ways than one. It was logical and carefully constructed, just like Casado’s writing, but it did not attempt to achieve any sort of balance. Instead, given the venue, it was strikingly partisan and tendentious.
Despite the technical window-dressing, it was devised to differentiate and distinguish Nicira’s approach to network virtualization from those of other players in the space, established vendors and startups alike. It also sought, implicitly if not explicitly, to derogate OpenFlow in the still-unfolding SDN hierarchy of value.
Just to summarize, though I encourage you to read the post yourself, Lambeth argues that, while there’s industry consensus on the desirability of network virtualization, there’s a significant difference of opinion on how it should be achieved. Network virtualization is not at all the same as server virtualization, he writes, citing the need in the former for “scale (lots of it) and distributed state consistency.” He concludes by saying that the current preoccupation with the data path, the realm of OpenFlow, is akin to “worrying about a trivial component of an otherwise enormously challenging problem.”
Positioning and Differentiation
Commenting on Lambeth’s post, Chris Hoff, formerly of Cisco and now with Juniper Networks (and a prolific tweeter, I might add), concluded correctly that it “smacks of positioning against both OpenFlow as well as other network virtualization startups.”
In issuing that positioning statement, Nicira not only is attempting to distance itself from the OpenFlow crowd; it also has at least a couple specific vendors in mind.
One obvious target is Big Switch Networks. If you visit that vendor’s website, you will find that it expresses unqualified love for OpenFlow on its home page. It also says candidly that “networking needs a VMware.” Diametrically opposing that view, Nicira says networking doesn’t need a VMware. Furthermore, as I noted in a previous post, Nicira continues to expend considerable effort to downplay the significance of OpenFlow.
Thinking Beyond Big Switch
But Nicira is thinking about competitors other than Big Switch, too. Readers of this blog will know that of one of my recurring themes — some would call it a conspiracy theory — is that the VCE partnership between Cisco and EMC is subject to increasing strain and tension. In short, EMC acquired VMware, Cisco didn’t, and now virtualization — and maybe VWware — is becoming integral to the future of networking.
Nicira’s Lambeth, formerly involved with distributed switching at VMware, and his counterparts at Big Switch agree that network virtualization is important. Where they disagree, perhaps, is in how it should be achieved.
Meanwhile, both vendors at one time or another, as Lambeth concedes at the outset of his post, have espoused variations on the claim that “networking needs a VMware.” Apparently, the team at Nicira has reconsidered that premise and is going in a different direction.
It might have adjusted course for reasons other than (or in addition to) those relating to architecture and technological requirements.
VMware’s Networking Ambitions
You see, VMware seems to believe that networking already has a VMware, whose name, conveniently enough, is VMware. Circumstantial evidence, including a recent post by VMware CTO Steve Herrod, suggests that VMware has ambitions that extend beyond server virtualization and well into network virtualization. Back in June, Greg Ferro also noted VMware’s interest in carving out a significant role for itself in network virtualization. In his commentary, Ferro cited a post by Allwyn Sequeira, security CTO at VMWare.
Herrod has predicted that “software-defined networking will become a mainstay of data- center architectures” in 2012. It’s safe to assume that he foresees his company playing a major part in making his prognostication a reality.