Automating network optimization has become very much in vogue, particularly as operators continue to look for cost savings and increased efficiencies. Growing deployment of Self-Optimization Networks (SON) has been driven by the promise of networks that can be the best they can be with very little engineering input.
However, not everything has been sunshine and roses. Many operators who implemented SON processes have found that they have to take the time to review the changes the SON systems have performed and ensure they are valid. In essence, automation still requires continuous monitoring to ensure the system does not break anything.
A key reason for this is that networks and user behavior patterns are not static. For their part, networks grow and evolve over time, often becoming more complex. Patterns of behavior can be different depending on usage across various geographic regions. For example, user behavior in New York City may differ from that in Topeka, Kansas. In these cases, a “one size fits all” approach can be detrimental, and automation highly inefficient.
For network optimization automation to be efficient and effective, operators can’t simply set their network optimization to autopilot – which is precisely what the word “automating” suggests. Automated actions must be accompanied and complemented by diligent operators who keep a watchful eye on their automation processes, or those processes will keep the network from achieving its full potential.
Left to it’s own devices, automated network optimization can be like whistling into the wind – operators may be doing something, but that something is having very little positive impact. That’s because successful and true optimization requires the ability to monitor and take action on the data that automation outputs. The machinery is only as intelligent as the people controlling it; ultimately, those people need to be the ones to interpret the data and put it to use. This requires a manual level of control that can be achieved by adopting the following strategies:
1. Put metrics in place before automating network optimization
This is the single biggest oversight we see, and yet it’s one of the most important things operators can do when optimizing their networks. Before automating their network optimization, operators should put clear and concise metrics in place to give them a good idea of where things stand before the process takes place. They can then compare the data they receive to those metrics to see where changes and adjustments need to be made within the network.
2. Closely monitor data output
Don’t just set the automation to “run” and let it go. Closely monitor the data output. See how many changes the automation process is implementing to get a better idea of potential problem areas, and proactively address those areas.
3. Continue to validate the data involved in the process
The old axiom “garbage in, garbage out” is very true when it comes to automating network optimization. As automation becomes more complicated, operators must continue to stay on top of the data that’s being put in to make sure it’s still valid. All data that’s being considered must be 100 percent correct or the results and accompanying adjustments will be inaccurate, requiring further changes down the line.
Still, done properly, automated network optimization can be a very good thing. It can make networks run cleaner so that they’re able to better deliver data and voice services. It can also free up time for engineers to work on other things that will help drive the network forward.
At the end of the day, though, automation is only as good as the metrics, processes, and techniques put in place to monitor it. There are different techniques that can be deployed for different use cases and needs, but they must be accompanied by baseline metrics and ongoing monitoring processes.
— Shervin Gerami
You might also like: