Cisco plans to predict, solve network problems via SaaS

Two years after it was it was acquire by Cisco, ThousandEyes’ intelligent network software is now at the heart of the networking giant’s strategy to predict and fix network problems before they cause trouble.

When it bought the compnay, Cisco acquired its cloud-based software package that analyzed everything from performance of local and wide-area networks to ISP, cloud, and collaboration-application performance to the health of the internet. 

Its cloud, enterprise, and endpoint-agent technology has been integrated in one fashion or another across Cisco’s core product lines such as its Catalyst, Nexus switches, and now that technology is part of Cisco’s latest effort: predictive networking.

Earlier this summer Cisco took the wraps off the first step in that plan with the ThousandEyes WAN Insights service that can provide Cisco SD-WAN customers with network forecasts and SD-WAN policy recommendations to improve application performance and user experience.

WAN Insights features a predictive analytics engine that Cisco says it will deliver via software-as-a-service (SaaS) to help organizations predict network problems before they happen.

The service is part of a variety of predictive services Cisco expects to offer that lean heavily on AI/ML and analytics to predict problems and automate their resolution across internet, cloud, and SaaS networking environments.

“There is more and more of a reliance on the internet as the corporate backbone, and the ability to forecast where problems might be [and that] is what we’re focusing on,” said Mohit Lad, general manager of Cisco ThousandEyes. “And SD-WAN fits nicely into that because it’s also very internet centric.”

Without requiring additional agents, WAN Insights applies statistical models to the Cisco SD-WAN’s vAnalytics path metrics to provide network forecasts and SD-WAN policy recommendations that optimize user experience across enterprise sites.

The package watches each network path and can analyze performance across all SD-WAN network sites to identify quality issues for multiple application categories. It can then can recommend alternate paths, at a circuit level, to better handle traffic. These recommendations are given for each site and application category, using quality estimates that reflect the requirements of each category, according to Cisco.

“Now when you look at your own SD-WAN you have the option of multiple circuits that you can choose from, and you have to be able to do better than just react to an outage. There are patterns and predictability [so] you can now plan for problems,” Lad said.

For example, if performance of Office 365 is predicted to suffer at 9 a.m. on a Wednesday, the service can recommend another link that might provide 20% better performance, he said.

“We are being very deliberate about when we say ‘predictive’ because we can’t predict every outage, but we can often-times can spot problems that customers really struggle with,” Lad said.

So a customer might notice a disruption to services every Monday at a certain time, but they don’t know the cause. “Those are the kind of problems that we want to solve with predictive technology, the ones that are extremely difficult to forecast and extremely difficult to fix but are the most annoying for organizations,” Lad said.

The more ThousandEyes is deployed in Cisco environments and elsewhere, the more widely it collects intelligence and the better predictions it can make over time, Lad said. “We built some original technology to be smart about what we can infer from the data we collect,” Lad said.

Visualization is also extremely important, Lad said. Network topology views can help interpret the data more meaninfully, for instance.

All of this work is really just beginning, Lad said. The immediate goal is to learn about what predictions enterprises do and don’t care about. Next the focus will be on driving more automation and incorporating data sets gathered by Cisco data-center products such as Catalyst switches, he said.