Open source communities are no doubt playing a key role in moving the telecommunications industry forward, but not everyone is on board the bandwagon.
Over the past five months or so, we've spent a fair amount of time writing about open source groups and standards development organizations (SDOs) such as the Linux Foundation, MEF, Open Networking Foundation, OpenDaylight, the TM Forum and ETSI, and there's clearly more cooperation afoot for the good of the industry.
But artificial intelligence startup B.Yond's chief product officer, Rikard Kjellberg, said his company has to be careful when it comes to choosing which open source community to commit its resources to. Kjellberg spoke to FierceTelecom on the heels of the AT&T Spark conference earlier this month.
"In general on open source, I think we're seeing some challenges," he said. "It was very evident at Spark, by the way. What's happening in Linux Foundation is that there's five or six different projects addressing edge cloud one way or another, and there's a ton of overlap and conflict of interest, et cetera.
"This was actually brought up at the Akraino meet up in Portland that I attended as well, where a lot of us who were there were struggling to make sense of this. Basically, Linux Foundation's position is that, 'Oh, this will sort itself out over time.' But in the meantime, companies large and small are investing and trying to place bets."
Leveraging seed code from AT&T, Akraino was designed to enable the scaling of edge services and to complement the Open Networking Automation Platform (ONAP) that is part of the Linux Foundation's LF Networking Fund (LNF.) At Spark, AT&T's Mazin Gilbert outlined how Akraino, Acumos and ONAP were designed to feed off of each other.
But where Gilbert sees a blueprint, B.Yond's Kjellberg sees Etch a Sketch drawings. Kjellberg pointed to AT&T putting roughly 6 million of lines of code from its internal ECOMP platform into ONAP as one example of the difficulties of open source.
"The next point is that when hundreds of companies collaborate, you are supposed to get accelerated innovation. No. You don't," Kjellberg said." I would say the approach that AT&T is taking is a challenge, given that AT&T is not a software company. What they've done, hopefully, with Akraino and Acumos, is that they have outsourced the development of that code. They submit millions of lines of code to the Linux Foundation and say, 'Hey, here's our gift to the open source community. Now you go build it.'
"And companies like ours are looking at looking at this going like, 'Oh, Hell, no. I'm not going to invest in this, because it's just not done right.' It's a beast. You have these tools that are supposed to manage cloud, and they're not cloud-native themselves. The code is not well documented. It isn't really tested."
Kjellberg said Google and Netflix are examples of providing fully mature open source projects.
"Netflix contributed Spinnaker. Works like a charm," he said. "Google contributed a bunch of things, including TensorFlow, which is an AI tool set. Works like a charm, because they used this in their production for a long time until they get to a point where they believe that, 'This is no longer core to us. Let's see if we can get others to contribute, and that will be our gift to open source.' And now TensorFlow has a life of its own and it's working great. Spinnaker, same thing
"That's really the model that works on the open source front. I think we're going to see a fair amount of challenges and churn before open source becomes mainstream in telco, but I do applaud AT&T's efforts in trying to build more infrastructure on open source."
During its Beijing software release this summer, ONAP made a point of saying that it was a deployment-ready platform. In addition to AT&T, which has used ECOMP internally for years, Bell Canada, China Mobile, China Telecom, Orange, Verizon and Vodafone have either deployed ONAP or are in trials to do so. Thanks mainly to ONAP, the LF Networking (LFN) currently has more than 100 organizations, which now enable close to 70% of the world’s mobile subscribers.
Earlier this year, AT&T worked with SK Telecom, Intel and the OpenStack Foundation to launch a new open infrastructure project for clouds and AI called Airship. As it did with ONAP, AT&T contributed its code to Airship.
"Airship is overlapping with ONAP," Kjellberg said. "There's a ton of overlap there. Now AT&T is basically making this Airship, which is the deployment and lifecycle management of workloads, part of their edge cloud Akraino. These are the things that need to be ironed out. All I'm saying is that I don't feel like Linux Foundation is helping there. They're cheering everybody on, but it's going to be ugly for awhile before these real reference architectures and real blueprints for how to do open source at the edge in a telco environment becomes clear.
"B.Yond is a small company. We're really careful with where we invest and where we invest our time. In some cases, we're making decisions basically saying that, okay, we could either place a bet on one of these projects and try to work with the semblance of a community, to the extent there is one, or we can actually make this functionality work using other open source and proven technologies."
Kjellberg said B.Yond is more focused on interfaces and not code, which is why it has adopted Kubeflow, which was put into open source by Google.
"We are adopting Kubeflow because we know it works," he said. "And so we're placing bets like that, because right now it's a time-to-market challenge for us, and we need to accelerate the implementation of the AI side of the platform. A lot of these things would've been a lot easier if people focused on the interfaces and looked at it as an integration challenge, not as a software development challenge."
Whether it has to do with open source or not, all startups need to make choice regarding their direction and the best use of their resources, according to Red Hat's Brian Gracely, director of product strategy.
"If you're a startup, you've got the normal problem of 'I've got a finite amount of resources, where do I apply them?'" Gracely said. "The foundations for the most part have gotten pretty good about not being an incubator or governance model for everything. I don't think that's as big a problem as somebody might perceive it to be, that there are so many choices. I think those choices are going to be there whether you're looking at picking a technology or you're picking which market segment to go after." — Mike
Editor's Corners are opinion columns written by a member of the Fierce editorial team. They are edited for balance and accuracy.