OPNFV evolves and hitches its wagon to Common NFVi Telco Taskforce

Birthday plans on are hold for the Common NFVi Telco Taskforce (CNTT) due to the coronavirus shutting down conferences. CNTT got off the ground almost exactly a year ago at the Open Networking Summit conference as a means to align the industry around unified network functions virtualization infrastructure (NFVi) implementations in order to reduce the friction for onboarding virtual network functions (VNFs) and, eventually, container network functions (CNFs.)

RELATED: GSMA and Linux Foundation join forces on aligning NFVi

CNTT, which is jointly under the auspices of the Linux Foundation's LF Networking and GSMA, works in tandem with the Linux Foundation's OPNFV Verification Program (OVP), which started out last year with the task of testing commercial VNFs against the reference architectures that the CNTT is developing.

RELATED: Common NFVi Telco Task Force publishes reference model and reference architecture for unified framework

The CNTT published its first reference model and reference architecture for a unified framework during its September ONS Summit in Antwerp. Beth Cohen, cloud technology specialist at Verizon, said the CNTT will publish its next architecture in June, after originally planning to release it at this month's Open Networking Edge Summit in Los Angeles, California, which was formerly the Open Networking Summit, prior to that event being cancelled due to the coronavirus pandemic.  The next release is called "Baldy" in honor of Mt. Baldy, which is a mountain California.

"Our theme is around mountains, near the places that we have the meetings," Cohen said. "We decided to keep the name because we liked it. The ONES conference itself is now going to be remote workshops. Then we're planning another release for the ONES conference in Antwerp in the fall if all goes well and the coronavirus curves continue to flatten out."

CNTT's first reference architecture, RA-1, was based on OpenStack and was announced at the Antwerp conference. The second reference architecture, RA-2, was designed for containers and Kubernetes. That reference architecture was fleshed out earlier this year. Cohen said a third stream, the edge, would work within the two reference architectures.

"We feel that edge is a superset of the other reference architectures," Cohen said. "So we've started work on the reference implementations and we're pretty far along on creating test beds. We're working with OPNFV so there's definitely tighter alignment with the OPNFV group."

In its latest release, which is called Iurya, OPNVF started to align itself more closely with CNTT, which was based on feedback from CNTT's members. AT&T's Ryan van Wyk, vice president of network cloud, said the decisin to have OPNV work more closely with CNTT came about during an LF Networking Forum in Prague, Czech Republic, in January that also included participation from LF Networking's Open Networking Automation Platform (ONAP) members.

CNTT embraces Kubernetes

 Van Wyk said that in Prague a much needed consensus was formed for RA-2 on items such as which kernel they were using and what version of OvS-DPDK, as well as defining boundary configurations.

"These types of things matter a lot," he said. "I think if you can get those things defined well in your reference implementation, then anybody can replicate that or it could be any distribution of OpenStack, any distribution of Linux, Red Hat, Mirantis or whatever. It doesn't matter. It's those implementation choices that really define the functionality that is possible within that implementation.

"Then I want to make sure that I'm conformant to that implementation. So that's where we get into the OPNFV testing, what they call their ODP testing ecosystem, as a way to make that sure that that implementation is conformant to that set of standards. You can have more than one reference implementation. So we may have one that's for large footprints. We may have one for more edge type footprints. The key is that you've got those types of implementation choices so when you put the workflows in you can be confident it's going to work."

With CNTT's scop3, vendors won't have to develop commercial VNFs for each service provider, and, on the flip side, operators won't have to spend as much time onboarding each vendor's unique VNF. Operators also won't have to alter their NFVi to get VNFs to run correctly.

The end-goal is to simplify and remove the unknowns so that deployments can accelerate in the software-defined networking (SDN) space, according to van Wyk.

Van Wyk said it was less about the NFVi conformance at this point and more about kicking the tires on the test model itself in order to have "something to throw at it so we can work on refining and improving it,"

"I think the other thing is now there's alignment around OPNFV and the role they play going forward as they become the performance testing arm of this taskforce," van Wyk said. "That was supposed to be where we would handle the badging if something is conformant to the to the reference implementation. So there's alignment with OPNFV on that being their role and they are supportive of it."

Cohen said OPNFV, which was launched six years ago, was pretty far along in the creation of the test beds that are needed to test the CNTT reference implementations against.

"There's definitely tighter (CNTT) alignment with the OPNFV group," Cohen said. "We're piggybacking on the test suites and a lot of the work that was done prior to continue the CNTT work going forward. We are also working on a badging program, and figuring out what exactly what that will look like. I know there'll be more conversations about that in the upcoming workshops."

Kirksey said the tighter integration with CNTT included several areas. The first was better integration with OPNFV, which includes getting the OpenStack RA-1 better integrated on the implementation side of the OPNFV requirements and specifications.

"We're now on the reference implementation and reference conformance work there (on RA-1) and we have kicked off cloud native," Kirksey said. "The conference has been postponed until September. We did have two days of technical workshops, basically interactive developer meetings and we have turned those into a virtual event. So the conference we are postponing, but we're still going ahead with the developer workshops so that the developers can have that opportunity to get work done."

On the Kubernetes, cloud native and containers front, Kirksey said that work got down to brass tacks during the Prague event, which also included brainstorming with members of the Cloud Native Computing Foundation (CNCF.)

"Prague was really where we started the cloud native work," Kirksey said. "There's a huge amount of interest within telecoms because I know that many of them have said, and that includes us, that there's a mandate to embrace containers."

Is CNTT the right approach for NFV?

Some of the big names on CNTT's roster include AT&T, Vodafone, China Mobile, Telstra, Globe Telecom, Orange, KDDI, ZTE, Verizon, Deutsche Telekom and Bouygues Telecom. Clearly there's a lot of effort and brainpower going into CNTT as the industry looks to make NFV live up to its promise.

While some service providers, such as AT&T, Verizon and CenturyLink have deployed NFV, it has made networks more complex and harder to manage in some instances. Currently, CenturyLink is constructing its fourth NFV iteration.

There are some naysayers within in the telecom industry that believe NFV is the wrong approach. Tom Nolle, president and founder of consulting firm CIMI Corp., is still not convinced that CNTT's efforts on NFVi are warranted. Nolle has long been a critic of NFV in general.

"I think the CNTT stuff reflects an old-line device focus rather than a cloud focus," Nolle said in an email to FierceTelecom. "In the cloud, you don’t establish specific classes of resources in your pool, and limit software to one of the selections. You provide a set of characteristics for resources, for application needs, and you combine the two to make a selection.

"Kubernetes is full of that stuff. A device-centric view says that you buy little, mid-sized, and big boxes, and everything you deploy then has to be fit to one of those categories."

While this month's ONES Summit has been cancelled, the Linux Foundation is hosting a keynote webinar — “The State of Open Source Networking and Edge” — featuring Arpit Joshipura, general manager, Networking, Edge, & IOT on April 30. Joshipura is expected to provide updates on CNTT, the OPNFV Verification Program and other new projects and releases during the webinar.