How ISPs Used AI and Machine Learning to Manage the COVID-19 Internet Traffic Surge

internet traffic and speed concept

Internet traffic during the pandemic continues to plateau since it experienced its largest surge in March and April, when more people were working from home than ever before and students turned to distance learning to finish off their school year. Now, with more people returning to work sites each month and students out of school for the summer, traffic has stabilized. However, there is no telling what the future holds as the pandemic continues on and as schools wrestle with reopening decisions for the fall. But no matter what happens, America's cable internet service providers are ready to handle whatever comes their way. 

The billions of dollars that have gone into broadband investment and infrastructure over the past 20 years have enabled cable operators to build, monitor, and upgrade robust and reliable networks that can withstand the added strain that unforeseen events, like the COVID-19 pandemic, have on network performance. Notably, these investments also extend to developing and utilizing advanced technologies, like artificial intelligence, to optimize network performance in the smartest and most efficient ways possible. 

Artificial intelligence and machine-learning software played a huge role for Comcast when the ISP experienced its largest surge in internet traffic this past spring during COVID-19 stay-at-home orders. Comcast, which has a network accessible to more than 59 million U.S. homes across the country, reported that its suite of AI and machine learning tools enabled the operator to deliver fast and reliable service even in areas where internet traffic spiked as high as 60%. 

According to Comcast Xfinity Senior Vice President for Next-Generation Access Networks Elad Nafshi, the need to invest in the development of these tools over the past 10 years was essential to handle the new and evolving demands that internet traffic presents every year and to allow the ISP to gain more visibility into its network. And during COVID-19, the tools have been critical to mitigating and resolving issues with network performance before customers experience them. 

"These technologies were born out of need and a recognition that networks could not efficiently meet the complex and dynamic demands facing them by simply getting bigger. While it's important that we have built an expansive, fiber-dense network to carry the volume of traffic that we handle, it's also essential that we continually invest in advanced software-based tools to manage the complexity of modern traffic demand," wrote Nafshi.

Comcast's AI tools worked behind the scenes to manage all of these complexities that came with the surge in internet traffic in March and April. Comcast Octave platform, a new technology that had only been rolled out to part of Comcast's network when COVID-19 hit, is programmed to deliver an increase in speed and capacity when it detects that a modem isn't using all the bandwidth available to it. In order to deliver additional capacity in a short amount of time to other parts of the network, a team of Comcast engineers worked around the clock to speed up deployment of the technology, which was completed in a matter of weeks rather than months. Their hard work resulted in a 36% increase in capacity just when customers were largely working and learning remotely from home.

The Smart Network Platform, another Comcast technology, is a suite of software tools that uses automation to reduce the number of customer outages and the duration of them as well. These tools have the ability to scan the network and take thousands of measurements every hour in order to detect service issues quickly and efficiently. In fact, thanks to machine learning, these problems are now identified in a matter of five minutes or less - down from 90 minutes - before many Comcast customers even notice anything is wrong. 

Comcast is also investing in a more virtualized, cloud-based network architecture, which reduces the need for manual upgrades by moving certain functions - performed by large pieces of hardware - into the cloud. The concept entails transitioning those functions performed by pieces of hardware into software that run in Comcast's data centers, allowing the operator to innovate at a faster pace, to use its AI and machine learning tools to scan the health of the network, and to correct problems detected automatically. 

The efforts and investment put into building out these technologies go a long way towards preparing America's broadband networks for unprecedented events, and even for the future as the cable industry continues to work on bringing 10G speeds to households. America's ISPs are constantly thinking of innovative ways to add capacity to its networks and to deliver optimal network performance in the fastest, smartest, and most efficient ways possible. AI and machine learning are just one set of tools that are helping them to pave the way and stay ahead of customer demand, even as the COVID-19 pandemic surges on.