Marketing hype fuels AI predictive network troubleshooting

One of the latest trends in the enterprise network space is the use of AI and predictive analysis to automate the identification and remediation of network problems. Cisco, for example, has touted its "first-of-its-kind predictive analytics engine that will help IT teams prevent issues and elevate the user experience."

The concept behind these automated network troubleshooting and remediation tools is straightforward. Networking and infrastructure components transmit various health, flow and diagnostic information to a centralized collector. From there, AI is used to organize, analyze and baseline a network's health.

When network components or traffic flows deviate from these established baselines, administrators are first alerted to the issue. Network operations staff are then presented with information on where the problem originated and steps on how to fix it. In some platforms, remediation steps can be automated as well.

Why all the hype?

AI-driven network prediction tools are not a new concept. In fact, they have been around for some time. For network-centric platforms, they are commonly referred to as AIOps. And, for security-focused network analytics services, they commonly go by the name network detection and response (NDR).

If AIOps and NDR have been around for some time, the question must be asked: Why the hype now?

The biggest difference between today's AIOps market compared to a few years ago is we're seeing the results of major network and infrastructure vendor acquisitions. In the recent past, these companies have gobbled up the most promising AIOps and NDR startups and have now integrated them into their own ecosystems. We saw this with Juniper's acquisition of Mist, VMware's acquisition of Lastline and Nyansa, and Cisco's acquisition of AppDynamics, just to name a few.

Of course, big vendors have big marketing budgets, and this is the real reason why we're seeing an uptick in media buzz around this topic. It's not that major innovations have occurred in these platforms over the past 12 months. Rather, these companies simply have a much larger audience and advertising budget to tap into.

Tighter vendor integrations can be helpful

Enterprise IT shops commonly gravitate toward the biggest tech companies because they're comfortable in those ecosystems or expect to have superior customer support -- especially when integrating one infrastructure platform with another.

In the case of AIOps, tight integration is required for the product to work as network and network security equipment must be configured to stream telemetry data to an AIOps collection and analysis engine. Thus, from a customer perspective, businesses only have "one throat to choke" for AIOps platform support if they operate in a single-vendor environment.

AIOps is great -- but it's not for everyone

For large organizations with deep pockets, highly complex networks and expert-level staff, I contend that AIOps is a must-have. However, for businesses that do not fall into this category, the amount of effort to implement these tools can be tremendous. Thus, few businesses are financially and technically prepared for these platforms, and therefore, most businesses are unlikely to realize an ROI within a reasonable time frame.

For organizations that feel they're ready for predictive network analysis tools, the best time to implement them is when a new network is deployed or when significant architecture upgrades are needed. It's here where network-centric AI platforms can be carefully configured, accurately baselined and thoroughly tested. This will help to ensure the tools are properly calibrated prior to relying on them in a production environment.