The use of AI tools and techniques have helped cable ops improve and enhance their ability to monitor and optimize networks during a pandemic that's still driving usage upward.

Karen Brown, Principal Analyst, KL Brown Consulting LLC

January 14, 2021

7 Min Read
AI becomes a critical network tool for cable ops

There's good evidence that artificial intelligence (AI) has become a bigger reality for cable operators in the past year, as they have begun to expand the technology beyond customer care chatbots applications to include network monitoring and management duties.

While adoption among MSOs is still in the early stages, it's not hard to see how AI's talent for gathering and analyzing millions of data points per second can be put to use in cable networks, automating fault detection and traffic monitoring. And that, in turn, may speed the evolution toward self-aware networks able to automatically fix outages while stopping content pirates and hackers dead in their tracks.

One of the strongest AI champions has been Comcast, which has incorporated AI-based data analysis into its recently launched Octave network monitoring platform. Using AI, Octave can analyze real-time performance data from millions of gateways and modems, identify network problems and then fix the issues. It can even adjust an individual modem's modulation profile to fit local access connection conditions, allowing some modems to run at 256-QAM while others can zip along at 4096-QAM.

Comcast also has used AI to assess local network capacity needs, allowing it to install 100 Gbit/s links in record time. The Philadelphia-based MSO also has partnered with Cujo AI to incorporate the technology's traffic monitoring and analysis capabilities into xFi Advanced Security service provided free to some 20 million cable subscriber households. Cujo AI also has struck a partnership with Charter Communications to provide in-home network monitoring and security services.

In Comcast's recently released Xfinity Cyber Health Report, Cujo AI CTO Santeri Kangas noted that artificial intelligence allows the xFi Advanced Security system to stay one step ahead of cyber bad guys, by learning what is normal network traffic for any device in a customer's home network – and what isn't. It can then block the bad traffic at the customer's gateway, shielding all devices in the home network. Cujo AI now protects some 750 million devices, Kangas added.

"Even for brand new devices, it only takes about 24 hours for our system to analyze, understand, and profile them for monitoring going forward," he said. "So, with each device added, Cujo AI's 'brain' gets that much smarter about how to protect your connected home."

AI's expanding focus

Meanwhile, Cox Communications has turned to Cisco's Network Services Orchestrator and Business Process Automation tools to automate much of its network support systems, ranging from trouble ticket and inventory management to IP assignments. It also plans to use AI analytics to add closed-loop processes to govern machine-to-machine interactions. And within its Contour video service, Cox is using AI to power Zone-ify, an app that captures what viewers like to watch to create a custom on-demand programming channel.

The cable industry also is delving into expanded use of AI through SCTE/ISBE's Explorer initiative launched in early 2020. The Artificial Intelligence and Machine Learning group headed by Charter Communications Principal Engineer Srilial Weera is tasked with creating a library of best practices for use of AI and machine learning, sharing lessons from employee and user perspectives and aligning with other CableLabs and SCTE/ISBE efforts. In June 2020, IBM joined the Explorer AI group, marking the first company outside of the cable industry to do so.

"The industry is going through a dramatic transformation as it prepares for a different marketplace with different demands, and we are energized by this collaboration," said Steve Canepa, IBM's global industry managing director for telecommunications, media and entertainment. "As the network becomes a cloud platform, it will help drive innovative data-driven services and applications to bring value to both enterprises and consumers."

One such innovative service is already underway at SCTE, which is developing an AI-based system to detect packet flows that look like content piracy without having to dig into the IP addresses or the packet payload. The key lies in red flags that can be detected in IP flow durations, total packets in an IP flow, packet lengths and intervals between packets. Using these indicators, the nascent platform so far has been able to cut the false-positive detection rate in half to about 0.2%, increasing overall accuracy to about 97%.

AI challenges on the horizon

Still, there are challenges to wider use of AI within cable networks. Given the COVID-19 pandemic, AI technology adoption in cable network management during the past year has not been a matter of interest but rather commitment, said Gil Katz, Harmonic's senior vice president of cable access business operations. For cable operators, network capacity was the name of the game, and for some the focus was on day-to-day operations rather than adoption of new technology. But at the same time, the new network capacity demands drove a need for better network management and monitoring tools, to understand shifting customer data usage patterns during the pandemic. So while some Tier 1 operators – and even some Tier 3 providers – already are adopting AI-based network monitoring and management tools, some are standing pat, Katz said.

Regardless of the MSO provider's position on the AI adoption curve, the technology is already at work in Harmonic's CableOS Central network monitoring and management platform, supplying improved data on usage, performance and anomalies. For example, in traditional cable networks, signal-to-noise ratio (SNR) levels are sampled every 15 minutes, raising the possibility that monitoring systems can miss issues. With the AI capabilities of CableOS Central, SNR sampling occurs every second, so "you can be proactive rather than reactive, because then you can see problems and fix them before there are issues and you get customer complaints," Nitin Kumar, CableOS Central's principal architect at Harmonic, said.

Looking ahead, AI management will likely extend to include customer premise equipment, allowing operators to fine-tune cable modem. Individual modems often have operational quirks; an AI-based management system can "learn" these issues and automatically make changes to improve the device’s performance.

"We are at the point now where we can detect cable modems that require a reboot, or we can even identify which element of the cable modem is not in the right state, and we can recover that process," he added. "I see that happening in the next 12 months."

Indeed, a wide range of ISPs including cable operators are starting to use AI in network management, primarily within event notification, according to Josh Chessman, senior director in Gartner's IT Operations Management Group.

"At the ISP level, the benefits for the organization are significant because an outage could impact hundreds if not thousands of users, and there could be direct monetary costs to that," he said, adding that is particularly true for commercial customers tied to more demanding service-level agreements. "If I can minimize my downtime by 10% or 15%, then there are significant upsides because there are the potential monetary impacts."

But Chessman also thinks it will take time for many operators to trust an as-yet unproven technology. Making sure early-stage AI monitoring tools accurately detect network faults is a concern – and that is magnified in systems that not only detect but attempt to repair issue.

"If you are remediating at an endpoint such as a laptop, the blast radius is small," he noted.

"But if you are remediating at a network level and it takes down all of Boston, Massachusetts, it's a much bigger issue."

Nevertheless, Chessman thinks AI's ability to digest and analyze massive amounts of data holds promise for wider-reaching applications in network root-cause analysis, problem identification and troubleshooting. But he also thinks it’s going to take time, and while the proverbial self-driving network is in the future, "I think that it is still several years away from it even becoming mainstream."

Karen Brown, contributing analyst, Light Reading

About the Author(s)

Karen Brown

Principal Analyst, KL Brown Consulting LLC

Principal Analyst, KL Brown Consulting LLC

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like