The path to network automation is now clearer and more defined
Network automation is a fast-moving space, and the 'whens', and the 'hows' of how organisations are opting to embrace automation continues to become clearer and more defined. It's worth exploring what that means for NetOps teams.
When it comes to the 'when' - the timeline for enterprise adoption - Gartner recently shared its expectations:
"By 2026, 30% of enterprises will automate more than half of their network activities, an increase from under 10% in mid-2023," the analyst firm said. In the same timeframe, "50% of enterprises are expected to use AI functions to automate 'day 2' network operations," which aim to improve the reliability, availability and performance of software and the associated traffic running over the network.
As important as the adoption timeline conversation is, defining what it should cover is too: how to pick candidate network flows and processes for automation; where to draw the line on what that automation is allowed to handle; and what to do if an anomalous or 'edge' scenario is encountered.
High-fidelity data for informed decision-making
A logical first step to introducing network automation at the scale Gartner suggests is to have real-time visibility of the network and of each component that plays a role in its end-to-end assembly.
By now, NetOps teams are familiar with operating a globally distributed network construct: where some parts of the network are owned and under their direct control, while other parts are unowned and under third-party control, but all need to come together cohesively to assure the end-to-end delivery of digital services.
In the owned and unowned networking construct, collecting high-quality data on each component and point of interconnection enables NetOps teams to visualise and glean insight from their complex networks.
That high-fidelity data is the foundation for informed decision-making, whether by person or by machine. We know now that in one-third of enterprises, within the space of a year, machines will be making decisions for at least half of the expanse of the globally constructed network.
That is both necessary, and a big change. It's necessary because the expanse of the globally distributed network, with its potentially billions of data points, is already beyond what one engineer or even a team can process. But it's also a big change, because within a year, this represents a sizable shift on the level of oversight responsibility that automation is trusted to achieve.
The path to automating network activities
We've established that end-to-end visibility and high-fidelity data are foundational to understanding what a NetOps team can and can't automate. They're also critical to ensuring an automation performs as intended, despite everything the globally distributed network might throw at it.
At its most rudimentary level, network automation is useful in break-fix scenarios - where one or more signals points to the presence of a familiar failure or outage pattern, triggering an intervention and improving mean-time-to-recovery. This could be as simple as an automation that diverts traffic onto healthy infrastructure based on upstream issues detected on regular routes.
That's fine, but at the point, automation is impacting over half of network activities, as Gartner suggests it is on track to, it's likely to be more sophisticated and operating at additional layers of the network architecture.
It's also likely to evolve beyond being used just for break-fix to be used in problem avoidance, continuous optimisation and improvement use cases. Already, this evolution can be seen in the tracking of the imminent expiry of TLS certificates, automatically renewing them before an expired cert can cause customer-facing issues - a not-infrequent cause of outages.
Trust is a key part of this expansion. The track record of an automation or AI - its ability to perform as specified - is already a determining factor in its adoption. Historically, automated straight-through processing, with minimal human oversight, has not occurred due to potentially costly false positives occurring.
With higher quality data being used to feed, train and finetune network automations, misinterpretations are becoming fewer. Still, even the best NetOps teams know best-laid plans can go awry. A functional, tried-and-tested automation can still produce an unanticipated outcome during a 'perfect storm' of issues. As long as engineering oversight and trust is factored in, this should not dissuade NetOps from aiming for the Gartner benchmark of greater than 50% of network activities automated by 2026, or even going beyond it.
