I’ve blogged a lot over the last couple weeks on the transformation of service provider infrastructure that’s being driven by the economic imperative of monetization and the technical mechanisms of the cloud, SDN, and NFV. I’ve also noted that this transformation will surely impact network equipment, and that one such impact will be commoditization of transport/connection functionality. The obvious question is whether there’s anything left for the network guys. There is.
The best way to think of both cloud applications and cloud-based services is as a “omnipus”, an any-legged octopus-like organism that lives in the cloud data centers but extends its arms, in the form of services, outward to the network edge. Each arm represents a service relationship, hosted on a virtual network in the cloud. Since the users of the future (like those of the present) are looking for applications and services, the users will transform their vision of “being online” to “being connected to the omnipus”, to all the arms that represent what they do and what they want.
This creates a kind of bicameral vision of service delivery; everything is “delivered” in a connectivity sense to the major points of presence where access (mobile or wireline) is concentrated. The user then simply taps into the necessary omnipus legs. Connectivity is commoditized partly by the natural forces of declining revenue per bit and partly because the connectivity needed at the service level is directed at linking fairly static resource populations (cloud data centers) with fairly static metroPOPs. This isn’t a non-connective network but it’s a network without much need for a lot of adaptation and dynamism. This is the trend that makes SDN and cloud-directed networking feasible; the network of the future covers most of its route miles in a very structured and easily provisioned way.
Every omnipus arm has a presence in the edge, from the network side. Think of it as being a doorway with a turnstile on it. The goal of the new-age edge is to manipulate the turnstiles. That means providing users access to services and applications by simply allowing cross-transfer traffic to one or more omnipus arms. This looks very much like a hybrid of a BRAS, a firewall, and DPI, because sophisticated traffic inspection determines what omnipus arm something gets connected with, if it’s connected to anything at all.
My view is that the shift to an omnipus service model is inevitable, and that things like SDN and NFV are symptoms of technical change being driven by an increased understanding that the future network WON’T be like the present one. This is why I’m so hard on vendors for not looking into both concepts now; they need to support not the early groping of buyers for the transition to the future, but the future that buyers are seeking. To do that they need to be thinking ahead, and that starts by thinking the here and now, not just pushing boxes based on a paradigm that is clearly going away.
Alcatel-Lucent should see this paradigm shift as their real opportunity. They have plenty of valuable assets to leverage this new vision, but they’ve had those assets for years now and have somehow been unable to leverage them fully. This, more than too many underperforming product sectors, has been their problem. With a strong monetization position that’s linked to both cloud and SDN, Alcatel-Lucent could be singing the song the buyers want to hear. That harmony has just not worked out for them, and if they could fix that problem decisively no other changes would be required.
SDN is one of the things Alcatel-Lucent needs to embrace, and the fact that embracing SDN isn’t rocket science is demonstrated by Cyan’s announcement of its Blue Planet platform, a software platform that rides on top of vendor management tools for major vendors like Alcatel-Lucent, Cisco, and Juniper as well as about a dozen smaller vendors (not all devices are supported). The platform supports a series of Blue Planet applications, those from Cyan in direct mode and those from third parties through a northbound API. Details on functionality and interfaces wasn’t provided in their release, and there was no specific mention of cloud-compatible interfaces. There was also no specific mention of OpenFlow.
The point is that management-system orchestration of service behavior isn’t new; anyone who can provision multi-vendor networks could assert SDN support, in no small part because we don’t have a strict definition of what that is. I’m of the view that Cyan is heading for what I would classify as an SDN strategy but that the state of the product wouldn’t meet my criteria yet. But even this would be a step forward for Alcatel-Lucent or even for Ericsson, Juniper, and NSN. And there are other companies who do as much as Cyan but who also do OpenFlow (M2MI is the example that comes to mind). I think we can expect to see a number of companies announcing SDN or network virtualization shortly, and while most of them will really be announcing a movement of an existing management orchestration platform toward SDN support, it’s momentum and it’s a way into the press. Those who wait a month or so are going to be in the “me-too” stage. You can also expect that we’ll see some hardware evolution, first in the data center and then in the edge network, but the latter probably won’t happen this year.
In economic news, US elections and the risk of a deadlock pushing the US over the fiscal cliff is suppressing markets, or at least it has up to now. Friday wasn’t a good day but the volume wasn’t decisive, and today futures are showing a modest upswing. In Europe, Greece faces a number of critical parliamentary votes that will determine whether it can draw on aid, and answer the “in-or-out” question relative to EU membership. It’s not likely that if Greece passes the necessary measures they’ll end up leaving, as some in Europe have already commented.
