Cisco unveils new AI engine that can proactively predict network issues

Cisco wants to enable operators and enterprises to proactively manage their networks rather than reacting to problems after they occur. To that end, it took the wraps off a new cloud-based predictive artificial intelligence (AI) engine it plans to incorporate across its suite of software-as-a-service products later this year.

Longtime Cisco Fellow J.P. Vasseur told Fierce the system works by pulling in data from multiple telemetry points and analyzing those to learn and make forward-looking predictions. For instance, the technology can be used to flag an expected path break or degradation of performance on the network, he said.

The idea is to apply the AI technology across multiple kinds of connectivity – everything from WAN, SD-WAN, optical and hybrid networks – to help improve the user experience. Cisco plans to announce the AI system’s integration with initial products during its Cisco Live event in mid-June.

Vasseur said he first presented the idea for the project to Cisco CEO Chuck Robbins in 2019, who ultimately decided to fund it. Since then, it has been working to develop an algorithm that strikes the right balance between accuracy in detecting issues and scalability. Vasseur noted a key priority has been avoiding false positives in order to maintain user confidence.

“The way I designed the system with my team is to say I cannot predict everything, I just cannot. For example, a fiber cut, how do you want me to predict a fiber cut? A power outage I cannot either. But what I want is no false positives,” he explained. “If you start issuing false positives, then the trust will be lost very quickly.”

He added Cisco was uniquely positioned to build such a product given its insight into different domains through its ThousandEyes network monitoring business and Viptela SD-WAN offering. Data from such a diversity of sources is invaluable when developing an AI model, Vasseur explained.

Once it had a prototype in hand, the next mission was to scale the product out in the field to ensure it could work with multiple customers and across different types of networks. Vasseur said Cisco has been fine-tuning the system for about 18 months and has offered “a few dozen” customers the opportunity to trial it more recently.

Robbins said in a statement “early adopters are seeing major benefits saving them time and money. The industry has been waiting for secure, proactive networking and only Cisco can do it right.”

Back in 2019, Ciena’s Blue Planet division similarly tapped machine learning and closed-loop automation to help prevent network outages. Last year, Comcast Business teamed with Versa to roll out a tool for its SD-WAN service which uses AI to understand network conditions and proactively route traffic accordingly.