AT&T has migrated millions of its broadband gateways to a new container framework from the Prpl Foundation to power application management on those devices. This plan comes as the carrier is set to boost the reach of its broadband services.

This migration involved AT&T replacing a legacy container orchestration platform that the carrier found to be inefficient at supporting a growing pool of API-linked applications running on top of increasingly resourceful CPE devices. These CPEs were basically starting to resemble white-box platforms used in router and switch architectures.

AT&T had been using a container framework “that was widely used in the Linux community and had a well-defined orchestration system to install, upgrade, and manage the containers.” It used this system to deploy its first containerized application, which was its ActiveArmor Internet security product that the carrier said has grown to run “millions of active containers.”

But, that load also came with significant resource requirements, “utilizing memory and CPU resources on the broadband equipment even when a container wasn’t installed and active,” which limited how many applications AT&T could simultaneously run on those devices.

AT&T explained that this resource issue came to a head in 2022, driven by the initial container orchestration platform having been developed to run on simpler CPE architecture using a specific chipset. Newer devices began to use multi-core ARM processors that were more efficient at running multiple applications.

This led the carrier to investigate the Prpl Foundation and its Prplware platform, which included and operating system, mesh implementation for Wi-Fi, and lifecycle management (LCM) for container-based applications.

The Foundation itself was established in 2014, and grew out of Linux-based OpenWRT OS project. AT&T is one of dozens of Prpl Foundation supporter, which also includes fellow service providers Verizon, Vodafone, and Orange, vendors like Nokia, Broadcom, and Qualcomm, and hyperscaler Amazon Web Services (AWS).

Initial use of the Prpl platform showed “significant memory and CPU savings when using LCM for containerized applications, especially when there were not active containers installed,” a savings that it added had convinced the carrier to use LCM “even on gateways that ultimately will not get moved to Prplware software due to age or where it is in the product lifecycle.”

The Prpl platform’s LCM feature also helped AT&T update legacy CPEs from their existing container management platform to the Linux-based environment. AT&T created a workflow in its CPE management system that following an LCM software upgrade would then trigger an automatic reinstall of containers to the new platform. The carrier said it has converted more than 12 million broadband gateways with millions of active containers to the new platform.

Looking ahead, AT&T noted that greater adoption of the new management system could provide developers with a common platform for the creation and deployment of applications that can run across multiple carrier systems and open more application choices for consumers.

AT&T’s broadening broadband and virtualization ambitions

AT&T’s container platform migration plan comes on the heels of the operator announcing plans to acquire Lumen Technologies’ consumer fiber assets for $5.75 billion. That deal includes a network spanning more than 4 million locations across 11 states and serving approximately 1 million customers, and bolsters AT&T’s fiber plans that now call for the operator to reach 60 million locations by the end of 2030.

The migration also continues AT&T’s more than decade-long push toward virtualizing its core network infrastructure. Those efforts were conducted under the carrier’s “Domain 2.0” program, which was based on disaggregating physical network and software components to drive more innovation and choice for the carrier.

Igal Elbaz, SVP and network CTO at AT&T, recently noted that more than 80 percent of the carrier’s total network traffic was running on that disaggregated network core, or nearly 840 petabytes of data per day.