IBM/AT&T deal on secure private access to IBM clouds has been called the “jolt the cloud needs”, which implies that the big problem with the cloud is the security of access. It’s a nice story, and even logical, but it’s inconsistent with not only my surveys but with others I’ve seen. The big problem with the cloud is that the buyer can’t make the value proposition work. Almost 2/3rds of the cloud projects in my survey stall because they don’t meet cost/benefit goals, and in almost half the cases the problem is that the buyer can’t really quantify benefits or fully develop the costs.
So what IS the “jolt the cloud needs”? The answer is first a good and systematic way of assessing cloud costs and benefits, and second a framework of cloud services that maximizes the second while managing the first. Vanilla IaaS cloud services, when applied to core applications, will yield costs much higher than can be achieved in-house. We have to move beyond vanilla, and move beyond simply moving core applications, to do better. When we can do that, the cloud will have its jolt.
You could say the same sort of thing for SDN. Light Reading has a podcast (http://www.lightreading.com/document.asp?doc_id=225727&) that gives a series of SDN-oriented vendor spokespeople a chance to talk about what they think the big myth of SDN is. The consensus is that the bit myth is that SDN is OpenFlow or network virtualization, and I agree, but that doesn’t define what it IS, and of course you can’t easily adopt something you can’t even define. But there was a comment by a couple of the people that an abstraction was a key element of SDN.
But what abstraction, or at the least, what source for one? RAD did an interesting blog on SDN (http://raddata.blogspot.co.il/2012/10/sdn-backward-step-forward.html), and the key point I think the blog makes is that SDN may be one of those all-or-nothing things. If you look at OpenFlow and the narrow issue of eliminating adaptive behavior in favor of central control, you have a hard time creating a stable vision of a future network. Again, I agree. The blog goes on to suggest that the key to making SDN a step forward and not a step back in networking is the union of SDN with the cloud to create a model of service that at least passes “central control” of the network to something specific—the cloud. The source of SDN abstraction is the cloud. That’s the high-level value proposition missing in SDN.
We could say the network of the future, or SDN, or virtualization, will “have its jolt” when we get that cloud linkage right. Quantum defines an abstraction for a virtual network, but it’s far from complete and surprisingly it’s not accompanied by (or preceded by) an abstraction for the cloud service or application itself. I think the real point of the RAD blog is that we’re diving down to the implementation of SDNs without having any context to guide us. Sound like the cloud problem? And could it be that the lack of a precise view of where the cloud is going is one reason we don’t have a precise view of where SDNs fit? Virtual computing, virtual networking, and virtual resources all have a great risk of becoming “phantom” instead of “virtual” if we don’t have a model from which we can orchestrate real behaviors.
It’s also interesting to speculate on whether a combination of cloud abstraction and SDN abstraction might not be the way to view things like the Telefonica Digital/Telenor cooperation on APIs. Recall that the Wholesale Application Community folded up not long ago despite “vendor support”, suggesting that simply providing an API forum wasn’t enough. Telefonica Digital has a strategy for exposing assets (billing, notably) through APIs, which puts the focus on assets, then on exposure, and finally on APIs. That’s how operators have always seen it. Could a cloud model include APIs and assets, thus creating a programming framework? You could model the Telefonica stuff that way, though they don’t present it like that in public…yet.
Telefonica Digital has broken with traditional vendors on the cloud and in the service layer in general, and they’re going to go big here. The Telenor pact is only the beginning, and vendors should be reflecting on how much they’ve lost, and are going to continue to lose, because they wouldn’t step up.
Why? Because mobile is now falling. On the other side of this picture, EU mobile operators are generally expected to lose BIG bucks over the next three years and the story this AM is that Japan’s Softbank is in talks to buy all or part of Sprint Nextel. All of this points the mobile sector finally losing its steam, and that’s been the only area where carrier capex has been strong-ish. This will put major pressure on vendors, and for most it is simply too late to take the right path now. We’ll see M&A and then consolidation.