Cisco reports its earnings tonight, so you might wonder why I’d blog about Cisco before the Big Event. The answer is that I’m not blogging about what Cisco did in the last quarter, but what it’s facing in 2014 (I may get to that last quarter on my blog on Thursday!)
Chambers has been very clear in his statements that Cisco’s future lies in becoming the top IT company, meaning that Cisco wants to grow out beyond networking. Its UCS servers have in fact been very instrumental in placing Cisco in the forefront of the cloud revolution, and certainly they’ve helped Cisco expand its total addressable market (TAM). The challenge Cisco faces is that servers aren’t enough.
According to my surveys, Cisco’s UCS successes have been in the best possible areas—network-related server applications like cloud computing. That’s helped Cisco to gain market share in the server space by leveraging its network incumbency and also helped to sustain reasonable margins on its servers, no mean feat given that servers are commodities in the broad market. The thing is, Cisco’s aspirations commonly lead to the statement that it wants to be the “next IBM”, and taken literally that would be a bad thing. IBM is suffering now, suffering from loss of strategic engagement created in part by poor articulation and in part by the diminishing role of computer hardware. Dell gained market share on both IBM and HP, demonstrating that servers don’t have the old-time strategic shine. And servers, stripped of the Cisco cachet, are what UCS is.
I think that most of the smart strategists at Cisco—and there are many—realize that the “next IBM” path has to take Cisco through the cloud, not through the server. In fact, cloud computing anonymizes servers because it virtualizes them. However, “the cloud” can’t mean cloud stack software because the fact is that no new proprietary strategy has a ghost of a chance of competing with current cloud stacks. Surely the cloud can’t mean “applications” to Cisco; they’d have to buy a bunch of companies to be credible. So what does it mean?
One path, which Cisco has sort-of-taken with WebEx, is the “cloud-means-cloud-services” track. Cisco’s biggest customers (network operators, government, and big enterprises) are generally happy to buy software as a service, and the former group would be happy to resell cloud-based services from others too. The problem with cloud services is that margins on IaaS are very thin, PaaS is something for platform/OS providers to offer, and SaaS means you either have to own software or have a killer developer strategy. I think this approach could ice some cloud cake if Cisco could get the lower layers baked, but it’s not going to fill the pan on its own.
Another path would be to look at the cloud-to-network boundary. Software-defined networking is IMHO evolving into a two-layer architecture with a software connectivity framework on top and a hardware layer driven more by policy than by centralized controllers at the bottom. Cisco has done a number of SDN-related acquisitions but it’s not really articulating a cohesive two-layer strategy at this point. Might Cisco meld all its stuff into a new software-overlay model that’s designed from the first for this new two-layer SDN approach? That would be a smart move in some ways.
The only problem with it is that any virtual-overlay model of SDN tends to make the underlay part, the hardware, into a commodity even faster than before. If hardware is where you make your bucks, that’s not good news. I think it’s possible to meld the layers to create a whole greater than the sum of the parts, but I don’t know whether Cisco would be willing to take the risk…I don’t know that I would if I were in their place.
That leaves the last approach, which is to create a cloud platform from the network side, what I’ve called a “supercloud”. OS players have a lock on PaaS now because most PaaS is a step up the stack from bare metal, meaning bare server. Why not think about PaaS as a step above bare resources in a more general way? Extend both hosting and networking in a common model, with the features exposed as platform APIs?
This is a notion with some precedent; Alcatel-Lucent’s CloudBand is in fact a model like this, though the company doesn’t position it that way (at least in a public sense). And if you look at CloudBand’s architecture you find a cloud-plus-network framework in another way—the Carrier Platform as a Service is an overlay on the cloud that adds additional network functionality. That’s where Alcatel-Lucent proposes to host network functions virtualization, for example.
The fact that a competitor is doing something like this could be a stimulus for Cisco or a negative. Cisco likes to be a “fast follower” rather than a leader, letting others take the big risks, but they also don’t like to look like they’re embracing a competitor’s approach. Alcatel-Lucent doesn’t represent any major threat to Cisco’s core router/switching business (nobody does) but they are a force to be reckoned with in the carrier space.
What might bring Cisco’s ultimate plan out into the open is NFV, and in particular the proof-of-concept initiative that’s developing in the NFV ISG itself. PoCs aren’t supposed to be competitive, they’re supposed to validate or refine elements of the NFV specification process. But it’s hard to see how a highly visible and highly credible PoC from a big network vendor wouldn’t be a standard against which other vendors would be judged. Can Cisco keep its own approaches under wraps if somebody like Alcatel-Lucent jumps out and does something highly credible? And remember, CloudBand is already an NFV platform as well as a platform for SaaS.
Cisco needs to make a move here, and soon. There are only a few opportunities to grab onto a credible “supercloud” mission, and other little grasping fingers are already brushing the prize.