Network AIOps is software that reads the alarms and readings from all the devices in a network, groups the ones that belong to the same problem, and names the likely cause. AIOps stands for artificial intelligence for IT operations. In a network, it turns a hundred separate alarms into one incident that an engineer can act on.
Why does a network produce so many alarms? The devices in a network depend on each other. When one link fails, every device that used it notices, and each one raises its own alarm. Routers report that a neighbouring router has disappeared. Radio sites report that they have lost their connection to the core of the network. Monitoring tools report that the apps behind them are slow. The engineers in the network operations centre, the room where an operator's network is watched day and night, then see a screen full of red. Finding the one alarm that matters is the job called alert triage, and doing it by hand takes time.
Network AIOps does it in three steps. First, it collects telemetry, the data that network equipment reports about itself, from all suppliers' equipment. It also reads the network's map of which device connects to which. Second, it groups the alarms that started at about the same time on devices that are connected. Third, it uses machine learning to rank the likely causes. Machine learning means software that learns patterns from past data, instead of following fixed rules. It also learns what is normal for each reading, using anomaly detection. So a slow change can raise a warning before anything fails.
Imagine a telecom operator's network operations centre at 11:20 p.m. A road crew working near Nagpur has cut a fibre cable by accident. Within two minutes, 240 alarms arrive from 63 devices: routers in Nagpur and Wardha, 41 radio sites and two data centre links. Network AIOps groups all 240 alarms into one incident. It notices that every affected device sits behind one fibre link, and that the light on that link dropped to zero at 11:18. It names the cause: "loss of signal on the fibre between Nagpur and Wardha, likely a cable cut near Nagpur". It also shows that traffic has switched to a backup route, so subscribers see only slower speeds. The engineer sends one repair team to the right road, instead of checking 63 devices.
The software receives device alarms from infrastructure monitoring and network alarms from telecom service assurance. It sends its grouped incident to the AI SRE, if the company uses one, or to the engineers' ticket system. It can also add the network's side of an app problem to the observability platform, so that app engineers can see that the network was at fault.
How is it different from infrastructure monitoring? Infrastructure monitoring raises each alarm. Network AIOps decides which alarms belong together, and what started them.