AI和边缘计算步入电缆聚光灯

全文内容 Full Text

At SCTE TechExpo, Charter and Spectrum explained how they are repurposing space, power and cooling at hubs and headends to serve up AI and high-bandwidth, low-latency services at the edge. AI is everywhere, so it's no surprise that it's being woven into the operations of cable broadband networks as operators look to stay steps ahead of service impairments, automatically fix some of them and improve overall customer experience. Meanwhile, broadband customers are using AI at an accelerating rate, but AI-related data consumption is not yet having much of an impact on overall network usage. AI in cable network operations is becoming a bigger part of the picture as cable operators and other broadband service providers gain access to more real-time data from their networks, including at the edge, where they've begun to deploy a new generation of nodes, amps and modems equipped with AI-capable neural processing units. Comcast has deployed more than 500,000 "smart" amps for its ongoing DOCSIS 4.0 Full Duplex (FDX) deployment, said Elad Nafshi, the operator's EVP and chief network officer, during Wednesday's general session at SCTE Cable-Tec Expo in Atlanta. Pictured (left to right): Matt Anderson, global market lead, telecom industry, Google Cloud; Elad Nafshi, EVP and chief network officer, Comcast; Justin Colwell, EVP, technology strategy and innovation, Spectrum; and Mark Bridges, president and CTO, CableLabs. (Source: Jeff Baumgartner/Light Reading). But there needs to be a way to separate the noise and manage that data to determine the context of issues – such as a fiber cut – and how to act upon them. "You can get choked on data and not know what to do with it," Nafshi said. Agentic AI is helping Comcast rapidly sort through volumes of alarms and when necessary put them on a single ticket, he said. And the data being obtained and analyzed is constantly being reflowed into the system to train AI models, with a goal of improving analysis and recommendations. AI "is as confidently wrong as it is confidently right," Nafshi said. Today, root-cause analysis agents are over 90% accurate, a figure that Nafshi hopes will be getting close to 99% by the time of next year's show. Progress is happening elsewhere, too. Nafshi estimated that Comcast is now able to "deconflict" more than 50% of trouble tickets. "We're just getting started," he said. Areas in Comcast's footprint with FDX are seeing a 59% boost in quality-of-service improvements and a 21% reduction in outage times. AI usage by broadband customers is also on the rise and is expected to continue increasing . However, the current impact on overall network traffic from those customers is still relatively small. Speaking on the same panel, Justin Colwell, EVP of technology strategy and innovation at Spectrum, noted that more than half of his company's broadband customers use an AI agent of some kind. AI-related traffic is growing quickly, up 350% year-over-year, he said, but it still represents 0.1% of the downstream usage and 0.3% of upstream usage on the Spectrum network. That compound annual growth rate is expected to continue rising as adoption of AI-assisted cameras, physical AI (robots) and self-driving vehicles increases. And that proliferation will have a material impact on the upstream, Colwell predicted. Those trends are demonstrating that the move to symmetrical gigabit speeds with DOCSIS 4.0 "will pay off," he said. Spectrum and Comcast are both investing in AI and embarking on new business lines at the edge of the network – including at headends and hubs. Those sites are being equipped with compute and Nvidia chips, and they are being used to open up new lines of revenue. That's coming about as both operators virtualize the access network, a migration that is enabling them to free up and repurpose space, power and cooling. Here at the show, Spectrum touted some use cases for deploying compute and low-latency connectivity at the edge, including demos of new,