What happens when the networking plumbing pipes of enterprises are filled with intelligence derived from AI? Better protection against outages? Faster response times for threat incidents? How, and what else? Let’s find out
According to a McKinsey Report ‘the State of AI’ 2021, predictive service and intervention appeared (18 percent) among the most popular use-cases of AI adoption. As per Cisco Global Networking Trends Report 2022, 45 percent of IT leaders cited responding to disruptions as the top network challenge for 2021. If only networking was a step ahead of the things that can go wrong! If only networking could learn from its past mistakes and gaps; and spruce up for more confidence ahead! Looks like Cisco has taken up that challenge, and possibility, quite seriously. We met Mohit Lad, Co-founder & General Manager, ThousandEyes (now part of Cisco) at Cisco Live in Las Vegas this June, and got a peek into this ambitious breakthrough. Excerpts.
What spurred this idea? How did it take shape?
For years, we have been working on different AI models and algorithms to provide recommendations and root-cause analysis. The earlier version of this was in WiFi when the network started to sense particular telemetry and could control the network automatically. Over time, we have tried to take the next step and get to a point where we can predict what can go wrong. The pandemic was a pivot and it became useful to have eyes on end-point devices in hybrid work culture. We have come to a point where we can learn from our past behavior with SD-WAN data and be proactive and not reactive. The company has been building and testing predictive software engines over the past two years. Early customer trials have shown Cisco’s technology can predict issues with high accuracy, helping IT teams to drastically improve connected experiences.
What can it do?
You cannot run from outages, but it helps to spot them early on. This also helps to reduce technical debt and create better flagging mechanisms. To top it up It’s not just about recommendations but how these recommendations would have played out had they been picked- in some cases.
But can it help to handle curve balls – if it’s all about past data?
It’s like forecasting based on past lessons. It’s an engine that uses historical data so not much room for curve balls in this use case. However, for non-SD-WAN use-cases, there may be some curve balls.
How much of this ties in with Observability?
Observability is a broad term. Cisco has a full stack of Observability capabilities. We can address a variety of network personas. Observability is not needed as a practice, but if it’s present, it helps in better troubleshooting.
What about data itself being a limitation- whether in terms of volume or the one beyond your radar?
Data can become a limiting factor only if use-cases change. Thousand Eyes has a very good state of visibility into the organization with continuous measurement expertise. We also have the vision to expand to other SD-WAN technologies. It is quite cross-platform and vendor-agnostic – with a good view of what’s happening inside the network.
How does it work? How will this be offered?
Cisco Predictive Networks gathers data from a multitude of telemetry sources. Once integrated, it learns the patterns using a variety of models. It starts to predict user experience issues, providing problem-solving options. So now, customers can decide how far and wide they want to connect the engine throughout the network, giving them flexible options to expand as they need. It can help network operators prevent outages, safeguard against attacks, and elevate the user experience – all while helping to increase network performance. Cisco plans to deliver predictive technologies across its portfolio in integrated, easy-to-use SaaS offers.
Mohit Lad, Co-founder & GM, ThousandEyes (part of Cisco)
By Pratima H