Cisco CEO John Chambers recently gave interviews to a number of technology journalists. The main point Chambers sought to hammer home was that Cisco had successfully “reinvented” itself after suffering a significant setback during the economic downturn.
Although Chambers spoke about several product lines and technologies — with video, a perennial favorite of the Cisco CEO, again receiving lavish attention — I will devote this post to what Chambers had to say about Cisco’s response to software-defined networking (SDN).
Going Old School
Reading what was both on the lines and between them, I think it’s safe to infer that Cisco’s data-center switching strategy will turn on ASICs, hardware, software, and services. While Cisco says it is open to working with SDN — OpenFlow is a relatively minor consideration, so let’s put it aside here — Cisco’s conception of software-defined networking is unlikely to bear any similarity to what we’re seeing from startups such as Nicira Networks or Big Switch Networks.
In stating at the outset that ASICs will continue to play a prominent role in Cisco’s switching strategy, Chambers tips his hand on SDN. He’s basically telling us that Cisco, if it espouses SDN at all, will favor an old-school interpretation of the technology, with a distributed control plane running across a switch-centric architecture. Contrast that with the logically centralized, server-based control planes being promoted by the aforementioned startups.
It shouldn’t come as any surprise that CIsco would adopt a grudging, incremental approach to SDN. Even today, Cisco makes most of its money selling switches and routers at robust margins. Anything it can do to maintain perceived switch value — and ASICs historically have played a big part in fashioning that value-enhancing narrative — will be aggressively pursued by the company. Notwithstanding that it has become a server vendor with its Unified Computing System (UCS), Cisco doesn’t want to accelerate a shift toward a compute-centric model of programmable networking in which applications and controller software on servers play omniscient puppet master to dumb switches. Where’s the Cisco margin in that vision?
Passenger on the Bus
Yes, that was a rhetorical question, and it probably didn’t require the imposition of a question mark. Indeed, Cisco isn’t constituted to play a leading role in the sort of brave new SDN world that has arisen from the campuses of Stanford, UC Berkeley, and other halls of academia.
Cisco will argue that the customer, and the aggregation of customers that form the demand side of the market, will be the ultimate arbiter. In making that assertion, Cisco would be right, but let’s consider that a certain constituency of customers, represented by large cloud service providers, already is voting with its data-center wallets and its considerable industry influence to champion the new model of SDN, which helps it “get the network out of the way,” innovate with new services, and reduce operating and capital expenditures.
Yes, Chambers touts Cisco’s membership in the Open Networking Foundation (ONF), but he must know that Cisco rides that bus and doesn’t drive it. As we all know by now, representatives from the largest of service providers populate the ONF’s board of directors, set its strategic course, and guide its technological agenda. Like other networking vendors, Cisco is a bit player in the ONF. For a company so cocksure and accustomed to dominating markets and standards bodies, Cisco cannot be pleased at its relatively plebeian status within the ONF.
What’s more, the ONF has no desire to cede power to Cisco in the future. Dan Pitt, the ONF’s executive director, stated recently that “no vendors are allowed on the ONF board,” and that “vendors can be on (working) groups but not steer them.” So, barring a dramatic change, Cisco’s influence on the ONF’s direction, and hence on SDN as defined by the ONF, will be marginal.
Fortunately for Cisco, the ONF is focused primarily on the needs of its board members. While a growing proportion of applications and computing cycles are moving to the public cloud, the enterprise networking market remains prodigious. Indeed, a great many enterprise data centers have invested in Cisco network infrastructure and architecture. As before, it will be difficult to persuade them to take their business elsewhere.
So, Cisco will fight a rearguard action against SDN and against software-driven networks predicated on merchant silicon. Looking to its past for future guidance, Cisco will continue to wager heavily on its ASICs, perhaps with a new twist here and there. For the most part, though, it will be business as usual. And business, in the enterprise and among smaller service providers, still is pretty good for Cisco.
The company isn’t in danger of dropping off the technology map, not even close. But there is a threat on he horizon, and past performance doesn’t guarantee future results.
For that reason, Cisco probably should have two sets of strategic plans — one for the near to intermediate term (which seems to be the plan it already has), and another that looks further ahead. In that plan, Cisco should envision a worst-case scenario — at least from Cisco’s perspective — one in which the SDN tide sweeps along smaller cloud providers and then begins to make incursions into large enterprises.
Even though it might not play out that way, Cisco needs to prepare for it anyway. The company needs to tap into some of Andy Grove’s “healthy paranoia,” because the compute-centric model of SDN poses an existential threat to the core of Cisco’s business.
As such, Cisco needs a proactive contingency plan, featuring acquisitions as opposed to incremental in-house development. Perhaps Cisco already has such a plan. If it does, Chambers and his executive team will keep it under their hats. There’s no point turning a worst-case scenario into a self-fulfilling prophecy.
Timing, as the saying goes, is everything.