Aruba and the Worlds of AI, ML and Self-Optimizing Networks

By Keith Parsons, Contributor
Share Post

Cutting Through the AI Mist

In a not-so-subtle session named Cutting Through the AI Mist at this year’s ATM Digital conference highlighted many of the AI-driven solutions coming to the Aruba community.

Another term we will be hearing more about, in addition to the terms AI (Artificial Intelligence) and ML (Machine Learning), is Self-Optimizing Networks.

Today, in the world of networking and the wider world of IT, the terms AI and ML are everywhere. AI and ML are the ante to be in the game. Using these terms, sometimes for mere marketing efforts, doesn't take away from the fact that we are getting more and more complexity in our networks, both wired and wireless.

But the real goal of all this added information is merely to help network infrastructure operators meet their desired requirements easier, and hopefully, more efficiently. Sure, any vendor can claim its AI/ML solution allows for automatic diagnoses and solving of network issues. But the proof is in the actual data sets and results-driven answers any automated system can bring to the table.

At this year's ATM Digital, we heard from a variety of Aruba’s experts and program managers on the various ways they are taking big data and turning it into solutions that can be used by mere mortals (compared with data scientists).

Achieving Stable, Consistent Connectivity

What we want as network infrastructure operators is stable and consistent connectivity for our users. So to help in that goal, Aruba is leading with AI-Operated Self-Optimizing Network solutions. These complex systems' goals are to collect and learn from massive amounts of data points—turning them into mathematical algorithms and using those measurements to improve our wireless, wired and WAN networks.

Take, for example, the simple transmit power (Tx) function on an access point. In the very old days—back with autonomous APs—we would set this number, in dBm or mW, and let the AP do its thing. If we had some user complaints, we may go onsite and take a couple of measurements and then decide to increase or decrease the transmit power.

Over the years, we moved to lightweight APs and added radio resource algorithms into our control plane in physical controllers. But with things moving to cloud infrastructures, we can now meld data from all our APs, as well as collected data on client interactions, into a cohesive model that can now adjust, adapt, and readjust to find the appropriate transmit power setting to optimize client connectivity and minimize client issues automatically.

AP transmit power is merely a single metric. It’s only one of many such adjustable functions that wireless LAN engineers use to adapt, modify, and tune our networks for maximum performance. No longer do we need to make monthly or quarterly validation surveys of our spaces to ensure clients are receiving properly tuned RF. Systems and algorithms can allow the wireless LAN to "self-tune"—or to use Aruba's term—"Self-Optimize."

Now take those same functions that help in the RF space, and expand your thinking to broader network functions like security, IoT, and upper-layer services like DHCP and DNS. All of these can and should be allowed to "Self-Optimize" as efficiently and quickly as possible.

I do NOT believe this is going to, in any way, intrude on the role of network engineers. We are merely going to be working on more creative and fulfilling roles doing network architecture and planning and less time on the day-to-day humdrum worlds of tuning the minutia of our networks.

I, for one, am looking forward to the more beneficial results of smarter network infrastructures!

Watch ATM Digital sessions on AIOps, Unified Infrastructure, and Zero Trust Security. 

See more highlights from this year's ATM Digital.