The operation of IT networks and systems undergoes a major transformation because traditional network management (ITOps) can no longer cope with the increased complexity of IT systems and the huge amount of data as new services such as 5G, Internet of Things (IoT) and API have emerged. Therefore, a new platform needs to be introduced: this is AIOps.
AIOps (artificial intelligence for IT operations) is a multi-layered technology platform that automates and enhances IT operations through analytics and machine learning (ML). Please note, that automation does not mean that there will be no need network operations personnel, but – as we can see in other areas of artificial intelligence – AIOps will make the tasks easier, more efficient and less stressful.
In order to present the elements and operation of the AIOps platform, we use the model developed by Gartner:
AIOps is thus based on a platform that processes extensive and diverse IT data from different sources using machine learning. Its task is to transfer the data from different data silos to a unified, structured and easily accessible place, where the machine learning system is then able to find patterns on the large amount of data and make a decision based on it.
The process of AIOps:
Observe: monitoring network alarms, KPIs, traces, and anomalies
Engage: Assign execution to special events detected during monitoring
Act: automation of analysis, workflows, and documentation
Several companies already offer tools to support AIOps. IBM’s solution is interesting, which takes into account not only log and alarm information (i.e. time series data) and numerical data from other databases traditionally used for IT operations, but also “text inputs” e.g. it also observes the dialogue between people (mail, chat) to act upon an incident in the system.
In summary, AIOps is a paradigm shift in ITOps required by the high level complexity, huge amounts of data, a high degree of failure, and ever-faster troubleshooting and reconfiguration. As systems have now outgrown human scale, there is a need for help from big data / machine learning – but it is important to express that it is not a substitute but a help to humans.