Cable network undergoing a 'radical transformation,' CableLabs CTO says
CABLE NEXT-GEN EUROPE DIGITAL SYMPOSIUM – Cable's hybrid fiber/coax (HFC) network is in the midst of a "radical transformation" that will add computing power to its baseline connectivity and be held together with virtualization, a top CableLabs exec said.
During his keynote today as part of the Light Reading online event, Belal Hamzeh, CTO and SVP at CableLabs, said the evolution of the network and a rethinking of HFC are necessary to prepare the cable industry to support a wave of new requirements for a broader set of high-capacity, low-latency applications. These next-gen applications will span everything from augmented and virtual reality and remote healthcare to mobile backhaul and edge computing and others that are still being thought of.
"The requirements of the network are becoming quite diverse," he said. "For us to efficiently and effectively handle that, we have to look at entirely new perspectives … Rather than looking at the platform as a connectivity platform, we need to start looking at the platform as a connectivity and compute platform."
Virtualization, Hamzeh added, is a "huge enabler for this transformation."
To tackle that critical piece, CableLabs has teamed with partners, including Altran, on an open source project nicknamed "Adrenaline" that aims to provide a centrally managed distributed and heterogeneous computing platform that supports the deployment of workloads across the operator's infrastructure.
The project also envisions a more agnostic approach to the compute architecture. In addition to running virtualization and compute on servers with x86 ARM processors, the idea is to expand support to a broader range of acceleration hardware, including FPGAs, DSPs, GPUs and CPUs.
This "technology-agnostic infrastructure" will diversify and expand how workloads are deployed on each of those environments and build in efficiencies in terms of power and speed, Hamzeh said.
Using Adrenaline, "the virtualization infrastructure extends from the regional office all the way down to the modem in the household," he explained, noting that the same stack can be used to handle massive workloads in regional data centers or much smaller workloads in consumer premises equipment (CPE). "With this new architecture, the whole network is transformed into a distributed, heterogeneous compute platform," he said. "You start looking at every element in your network as a location where you can run workloads on a various set of hardware."
For example, Hamzeh said this approach would allow an operator to migrate workloads from a hub to a headend or down to the CPE level based on the workload requirements, and that workloads could also be changed or shifted dynamically depending on the time of day. While some workloads are best handled in a hub or headend during peak times, they could likewise be later moved during non-peak times to FPGAs or CPUs and out into remote PHY or remote MACPHY devices to drive big power savings.
This kind of workload flexibility, "enables you to do dynamic workload shifting and optimize how the network is running," Hamzeh said.
Adrenaline has been scaled out and deployed as an open source. Today it can support current HFC infrastructure with x86 ARM-based virtualization, but it is in position now to begin adding acceleration hardware support on an as-needed basis.
"As with anything in the network, this is an evolution, not a revolution," Hamzeh said. "My expectation is the driver for this will be the applications, but the beauty of it is the network will be ready to support these applications."
Light Reading will provide more coverage of the Cable Next-Gen Europe Digital Symposium through the week. Day two of the event gets underway Thursday, June 11 at 2 p.m. British Summer Time. Registration information is available here.
- CableLabs, Altran team to take open source to the edge
- CableLabs Connects With Kubernetes
- Comcast's Virtual Access Network Rolls Into Multiple Markets, Spans 100K+ Customers
— Jeff Baumgartner, Senior Editor, Light Reading