ANOMALY DETECTION FOR THE MONITORING SERVICE
Anna Tsvetkov Kravchenko · Zenodo (CERN European Organization for Nuclear Research) · 2023
Monitoring has proved to be a crucial part of the operation lifecycle of any computer system, as it provides access to important internal information about the behavior and state of the system at a given point in time. With the evolution of the systems more complex monitoring cases emerged that need an innovative approach, achievable only by involving machine learning and anomaly detection techniques. The IT Monitoring Service at CERN (MONIT) is a central infrastructure collecting data and providing advanced dashboards and alarms for the CERN IT Data Centre and the WLCG infrastructure, and as any other computer system it also requires monitoring of its internal components in order to guarantee normal operation and avoid the potential outages. Aiming at covering the anomaly detection demand, the IT Monitoring in collaboration with other IT Computing teams have been working on a project (ADMON) meant to simplify the development and production deployment of AD models. The major goal of this project is identifying and creating an anomaly detection model for monitoring use cases within the MONIT infrastructure that cannot be easily covered by the traditional threshold-based alerting methods. It will also make use of the ADMON infrastructure in order to simplify the process and prove its efficienc