For communications service providers (CSPs) planning to deploy edge computing as part of their 5G strategy, the good news is that you won’t have to replace your existing NFV architecture – you can extend it to the edge. The bad news: there’s no standard way to do that.
During a fireside chat session on 5G and edge computing at last month’s Digital Transformation Asia in Kuala Lumpur, Vanessa Little, director of Global Telco Innovation at VMware, said that deploying edge computing nodes doesn’t require a whole new NFV architecture.
“When you really look at what’s under the hood there and how its infrastructure works, it’s the same architecture – just now we have some smaller data centers that are extended a little further out there that we need to connect into the central data center,” she said.
A key challenge, she said, is a distinct lack of standards for edge computing. “There’s no standardized APIs – we even use different terminology for the same thing. There’s a lot of people going a lot of different directions and no one’s really come together yet under common cause.”
Even the open-source community hasn’t yet managed to band together on edge computing, she added.
“Right now we have a lot of different competing projects that are popping up … but because there’s so many of them we have no de facto standard. It’s so open ended that it’s almost too open,” Little said. “Without some overarching standards to govern or give a little guidance and direction to these communities, they’re all over the place.”
The lack of standards is also making life difficult for developers hoping to write apps for edge computing, said Vinod Joseph, lead architect for the Office of the CTO APJ at VMware.
“If I’m going to build an application for cloud gaming as an example, when a user moves from Cell Site A to Cell Site B, they directly connect to another edge compute location. So how would the platform provider or a telco provide the context and the state from Location A to Location B so that I don’t lose the context that I, as a user, am connected to?” Joseph said. “As a developer, do I have to build that logic into the application, or is there a way for the platform provider or the telco to hand that to me?”
Vincent Seet, head of enterprise architecture at Globe Telecom, said that as a CSP he would love to be able to provide that info to apps developers, but there are no viable solutions commercially available. “There’s no platform, so to speak, and developers are still left on their own to figure this out, which makes the whole development process very difficult.”
Little of VMware agreed, adding that the platform and the RAN need to be able to share information for the app in near real time, but solutions for this problem are mostly proprietary and siloed.
Little stressed that there’s plenty of industry interest in collaborating on edge computing – what’s missing is a structured and credible way of doing it.
“Things like TM Forum offer great facilities and are excellent facilitators for things like this. The work that they’ve done in the open APIs is directly aligned with what we need to do for edge,” she said.
Seet of Globe Telecom agreed. “I think there’s some opportunity for the TM Forum folks to consider establishing some best practices or even a framework for edge computing so that there’s more structure and we can mature faster.”
Little added: “I would even go so far as to say let’s take it all the way down to the API layer – develop some standardized industry-driven APIs that we can all agree on, and develop against.”
This article was originally published at TM Forum’s Inform