The end game for any new technology is deployment based on business value, and I’ve noted many times that the biggest challenge faced by NFV is getting out of the lab and into the network. That’s now starting to happen, at least as a small step, through a Masergy service called Virtual f(n). It’s not yet a proof point for NFV overall, but it suggests that there are real new service revenues out there for NFV.
Masergy is a private company that’s specializing in cloud networking or perhaps better yet in NaaS. They’ve been well-regarded in the industry, and as an “alternative” provider of business services have always focused on agile deployment and service enhancement as a differentiator. Virtual f(n) is a continuation of that approach.
The foundation concept in Virtual f(n) is an agile vCPE box, provided by Overture Networks, in which network features can be hosted. In the initial deployment, Masergy is hosting a router from Brocade and a firewall from Fortinet. The Overture box is their 65vSE, which is based on Intel’s Atom CPU, and which Masergy says is one of the very few that can achieve line-rate performance in a service chain application.
The Masergy business model is a big driver for Virtual f(n). Their customers aren’t concentrated in a few metro areas but rather are scattered over a truly global geography, which makes it highly important to be able to start up and modify new installations without a lot of truck rolls. With the 65vSE, Masergy can load features into a premises device as needed, and couple these features with the highly agile cloud network service they offer overall. It seems pretty clear that Masergy intends to go beyond just two virtual functions, though it’s not yet clear just what their next steps will be.
This is an important step forward for NFV, obviously. Operators have been telling me for two years now that for NFV to roll out they’ll need benefits beyond the original notion of saving money by displacing appliances with COTS-hosted software. Proving those benefits hasn’t been easy, and in fact Masergy is the first true commercial NFV-based service deployment I know about.
There are two technical things about the Masergy deployment that I think are important. First, Masergy picked a company that has an open VNF hosting model. Overture has partners, of course, but they have worked hard to make the 65vSE an open device that’s essentially a locally sited piece of NFVI. Second, Overture is one of only four vendors that has what I believe to be a truly credible orchestration story as part of their Ensemble suite. I think both of these played a role in their selection, and I think that in the long run orchestration and management is what will make the Virtual f(n) concept operationally efficient enough to be profitable.
It would be nice to say that this is a giant step toward NFV deployment, but it wouldn’t be a fair comment. It’s an important step absolutely, and the vCPE-driven or edge-hosted model offers carriers a major advantage in controlling first costs and delivering benefits that scale with costs. It’s also obviously somewhat specialized to business services, and Masergy is most likely to benefit from it because of new customer turn-ons and changes in service topology. None of those factors would be enough to create a major NFV explosion, and none have a clear pathway to evolve toward a centralized, cloud-hosted, NFVI that’s still seen as the essential element in NFV future.
What I think this deal does is demonstrate a point I’d made a couple blogs back. We are now to the point where “standards” are not going to drive NFV, and where NFV PoCs are going to be of very limited value. There are too many things that have to be done in order to make NFV beneficial enough, and there’s precious little being done to get these things completed. Vendors, I’ve suggested, will have to take the lead and drive NFV forward by boldly supporting needs even if it means stepping outside the scope of traditional NFV standards.
Overture, for example, has developed a repository-and-analytics management framework that is much closer to the ideal for NFV than the ETSI vision of per-VNF managers. They’ve also promoted an edge-hosted, generalized-appliance, vision for VNF deployment rather than a centralized pool of resources. But they’ve done something useful, very useful.
The thing to do now, I think, is watch where Masergy takes Virtual f(n). I’m particularly interested in what happens with the Brocade/Vyatta router stuff. Right now, Masergy has a lot of real iron in its networks. If it takes SDN and NFV principles seriously it will start placing customer-network-specific cloud-hosted router instances inside the cloud at topologically convenient places. It will start to use Overture’s box to provide multiple Ethernet service virtual terminations for a variety of services. And if it does all of that it could be the first example of a real SDN/NFV network to create global commercial services.
It could also become the first to prove out the operations questions. vCPE has an advantage over “traditional NFV” in that it’s loading virtual functions into customer-located-and-dedicated devices. That simplifies management considerably by eliminating the shared-pool resource management issues that NFV would normally create. However, it’s very possible to see new VNFs getting deployed that are both more dynamic than edge routers and firewalls would be, and even to see Overtures’s orchestration being spread to aspects of interior network control. If that happens, then Overture might gain some significant experience in operational integration, something that’s absolutely critical for NFV.
So far, none of the players here are declaring this to be the Dawn of a New Age, which I like because it’s still a bit too early to say whether the lessons of Virtual f(n) will be broadly useful in promoting NFV to other operators and in other missions. However, this is an absolutely critical step from vendors (Brocade, Intel, and Overture) demonstrating that you can boldly go beyond the too-narrow scope of traditional NFV and find gold.