We’re in earnings season now and so we’ll likely have company results to comment on for a couple more weeks. Monday is a good day to consider the broader trends that were crowded out last week by news, particularly news of a financial nature, so that’s what I’ll do today.
One thing that’s developing at a systemic level is signs that the wireless service market is transforming. I blogged earlier this year that the consensus of Street estimates and my own survey/modeling activity was that the ARPU growth rate in wireless would go negative some time toward the end of 2012. There’s now information to suggest that the postpay wireless space, whose performance has been driven largely by smartphone contracts, is topping out and that even prepay may be nearing capacity. At one level this is certainly no surprise; at some point every market saturates and thereafter grows only as fast as its demographic base. The problem for networking is that it’s coming at a very bad time.
Global carrier sentiment about additional network buildout shows that future growth in spending would have to be tied to more direct proof of return on capital invested. Operators realize that return is easier to prove either up top of the service stack, where services live (on the revenue side) and OTTs compete, or down at the bottom where capacity is created (a cost-side assessment). The operator drive toward optical core networking as an alternative to hierarchical routing is an example of latter; a drive to create bits at lower cost because lower is where price-per-bit is heading. Mobile has historically been a sweet spot in a graying picture because ROI has been higher, so having ARPU plateau now is bad news. The idea that the smartphone revolution is going to stop driving more expensive plans is worse news.
The cost-drive portion of the shift in mobile spending should have framed the Open Networking Summit last week, but it didn’t. Software-defined networks (SDNs) and OpenFlow show great promise, but they have scalability limits that mean the technology has to be carefully matched to application requirements. The problem is that there’s no indication ONS is taking up the question of whether the special issues of wireless backhaul, for example, create a viable OpenFlow application. I think they do; at least the traffic plumbing of backhaul that’s now seen as Evolved Packet Core might be re-framed in OpenFlow terms. Is there a discussion here? Not that I’ve heard.
ONS turned out some interesting stories, but so far only one of them has been “revolutionary”. Google says it’s using OpenFlow on 100% of its internal backbone, and this should shout “OpenFlow is a Cloud Strategy!” to anyone with ears. So far it hasn’t seemed to have done that. In fact, I think that the stories out of the event show that getting associated with SDNs may be a higher priority than doing something actually useful. In short, we may be pursuing “softwashing” with as much determination as we’ve pursued cloudwashing.
Most of the SDN announcements being made, most of the demos being done, relate to the “support” of OpenFlow on switches or routers. I’m also seeing early attempts to map something like MPLS or GRE tunnels to “software” and call the result an SDN. That’s not the path to SDN or OpenFlow success, it’s a defense against OpenFlow. I think OpenFlow and SDN principles simplify networking, creating the notion that what’s allowed to connect is what’s supposed to connect. A simple, explicit, model. To add OpenFlow to a device that already supports universal connectivity isn’t going to cut it.
The optical space is important, maybe critical, to OpenFlow because it could bypass the standards and vendor inertia, as I’ve blogged before. The cloud space is also critical because it could link the SDN/OpenFlow concept of explicit connectivity to several models that would be aware of the explicit connection requirements. Get these right, get these linked, and you have OpenFlow success. Otherwise you’re just softwashing.
Who’s at least appearing to lead in OpenFlow and Optics? Not optical vendors, but Ericsson. They have substantive work being exhibited at the Summit, not just “I-too-do-OpenFlow” junk. They’ve presented a dozen good papers on OpenFlow/SDN topics, including some ideas on linking OpenFlow into IP networks. Ericsson, after all, has a lot to gain if I’m right about the future “cloudnet” looking like a bunch of BRASs surrounding a cloud-and-SDN core.
But Ericsson has nothing to position it in the cloud space, at least nothing I’ve been able to find. They joined OpenStack just a couple months ago, though, and in one sense it’s early for them to have a mature approach that combines OpenFlow, optics, and the cloud. On the other hand, they didn’t surely could have joined earlier (it’s not like they could have missed the need for a carrier cloud strategy), and the OpenFlow-to-cloud linkage is needed now regardless of what excuses might justify delay.
The cloud and OpenFlow also impact security, another issue that’s getting washed in the coverage of the topics. It’s very true that businesses want to see a unified security model for the cloud, as recent reports say. It’s also true that if you presume that (first) the cloud’s services are still being determined by market dynamic and (second) that OpenFlow will impact the cloud and security both as a collection and separately, you can see that there’s a lot of room for discussion. In an OpenFlow network, connection permission is explicit like a whitelist, which means that the processes that determine how the OpenFlow controller decides to permit them are the things that need to be secure. In the cloud, addressing and security could be fairly linked to the overall DevOps process, which as I’ve said can also be linked to OpenFlow…you get the picture.
I think that SDNs and OpenFlow are revolutionary, that they’ll be the de facto strategy within the cloud and the data center. Properly supported, I think OpenFlow is the solution to metro and backhaul, the EPC, and the OTN. I also think that there needs to be some insightful dialog on all the issues I’ve noted here, and likely more, to realize all of this good stuff. I wish we’d here more about that and less softwashing.