How smart was Nokia’s move to embrace Red Hat for platform software rather than continuing to try to support its own stuff? The move has been reported to be focusing on things like 5G Core, but I’m hearing stories that it’s a much broader and much more important step. If that’s true, then it’s fair to ask whether Nokia is about to seize the high ground against competitors, including Ericsson. The fact that of the 53 operators who offered me comments on the deal, 51 were “positive” and two were neutral.
For over a decade, network operators have been confronting a kind of service dualism. Connection services were still largely based on devices, with software embedded in the hardware, but higher-level services were targets of a shift to a server-hosted feature model. Network Functions Virtualization (NFV) codified an approach to that, but it had a problem from the first in accepting that feature hosting meant cloud computing, and that had to mean that NFV would build on cloud technology. NFV went its own way, and while it’s since attempted to retrench, it lost the opportunity to shape feature hosting principles.
Vendors like Nokia were part of the NFV work from the start, and perhaps that influenced them to view feature hosting in what was a less cloud-like way. They were also, as mobile infrastructure giants, used to providing a complete solution for the critical radio-access network (RAN) component of mobile networks, and that influenced their decision to push their own hosting platform tools rather than adopting a more general approach based on the stuff used to build cloud applications. Their decision to reverse that and buy into the Red Hat strategy has major implications, not only for Nokia but perhaps for NFV and for the market overall.
While, as I noted in opening, that the Nokia decision focuses on 5G Core, I think it’s pretty obvious that Nokia wouldn’t adopt Red Hat to provide platform tools in one area while rolling their own feature-hosting platform in another. What I hear is that Nokia never really intended to focus on 5G Core, but rather on replacing the Nokia Container Service that provided overall feature hosting. My sources tell me that the primary driver was the increasing interest pf operators in public cloud partnerships. Public cloud infrastructure obviously tracks cloud technology evolution, and having its own container service meant Nokia risked becoming incompatible with some public cloud hosting services and the apps developed for those services.
I think you can read this into a quote in the story I cited above. “Our CSP [communications service provider] customers have been telling us they want more choice and openness. More and more CSPs are making separate buying decisions on cloud infra and applications.” This implies a link between adopting a third-party cloud platform and “choice and openness”. That’s true if you extend the goal of feature hosting to a broad set of evolving services, since as I’ve noted in other blogs, telcos are actually embracing vendor solutions for 5G RAN, and they don’t seem to be demanding generic open-model software for 5G Core either.
What’s implied here is that operators want open platforms more than they want open applications. Why? What some are telling me is that their investment in features is really primarily at risk at the platform level. They spend more on the servers and platform software, by far, than on the 5G components or other prospective feature elements of new services. That’s the investment they need to protect, and a Nokia-specific (or any network-vendor-specific) strategy for feature hosting is a risk to that goal.
There’s still a “why this?” question, though. Obviously it would be possible to offer features, 5G Core and beyond, in a kind of “cloud-neutral” form. Operators, though, admit that they don’t necessarily understand how to assess the true neutrality of a feature offering. Operators also admit that they’d like to have a vendor offer the platform software as an element to their solution. Red Hat gives Nokia a portable platform that’s proven itself widely in the enterprise, in multiple missions.
But why now? Again quoting the article, “We’ve been very open and very deliberate for a couple of years now that our cloud infrastructure offering was really [only] designed for our applications,” he [Nokia’s Fran Heeran] says. “We were not in the market of selling a general-purpose horizontal cloud, hosting multiple applications from multiple vendors.” This was the NFV approach; one-off a platform for telco applications, and not only the NFV itself is moving away from it. So are telco-centric vendors like Nokia. The reason is those operator relationships with the public cloud. Operators are not going to deploy hosting out of the areas where they have real estate and staff to support it. They have to host stuff on public cloud services, period. Thus, they have to offer software features that run there, as well as on in-region operator hosting. Platforms like Red Hat are an obvious solution.
Ah, but “like Red Hat” implies there were other options available for Nokia, and the principal one was VMware. Why not them instead of Red Hat? There were three reasons offered me on this point. One was that VMware’s deal with Broadcom is still in flux, and operators in general are more positive about VMware in combination with Broadcom than they are in VMware alone. Broadcom introduces white-box options that could increase openness overall, but only if the deal goes through. Right now, VMware management may be wary of major deals that could raise further anti-trust concerns, and operators seem to think they could move faster with Red Hat.
Reason number two is that Broadcom may collide with Nokia’s own products and some of its management. Unlike Ericsson, Nokia has router products and more of a device stake in the future, with “scalable, secure, and adaptive” network devices. Nokia sells devices that white boxes would compete with. Broadcom’s white-box position, derived from its excellent network chipsets, could well create angst for Nokia’s router people.
The final reason for a Red Hat deal is private 5G and the public/private tension in edge computing. We have “edge computing” today in virtually every enterprise IoT application; 97% of enterprises tell me that. It’s based on their own local server or appliance technology, and it’s been gradually evolving to tie itself more closely to the public cloud. Private 5G is an application of this fluid edge computing notion, and operators know that they may want to offer it to enterprises rather than have them go elsewhere. Some already do. If you need a strategy that can support cloud and local hosting of 5G control plane elements, you need a generalized cloud platform. If, as an operator, you’re going to offer 5G feature hosting you may find IoT feature hosting just as important, and so you’ve justified an open, cloud-centric, platform model overall.
The net effect of all of this is to promote a model of “carrier cloud” that’s virtual in itself. Carrier cloud isn’t operator-specific platform technology and dedicated servers in operator facilities, it’s a set of applications that can be spread around, just as enterprise applications are, between different hosting locations and administrations. Operators seem to be gravitating toward a view of carrier cloud that’s disconnected from server purchases and major data center build-outs. This view doesn’t preclude any of that, but it doesn’t demand it either.
What would other network vendors do, though? Ericsson clearly has to make some move here; do they follow Nokia with Red Hat? I’d suspect that their management would prefer to make a differentiating choice, and unlike Nokia who could find a Broadcom/VMware deal problematic, Ericsson might see it as a way of using Nokia’s device position against it in a battle for openness. Cisco, a network vendor who offers servers and platform software tools, may also have to respond to the Nokia move, given my view it’s not limited to 5G Core. Even network vendors who don’t field their own platform products may need to think about formalizing a relationship in the hosting platform space. The network of the future is going to be increasingly dependent on hosted features.
