ethodology and classification of open-source ML methods for IT monitoring based on the Zabbix system

Abstract

In this paper examined the use of open-source machine learning methods for IT monitoring tasks based on the Zabbix system. Analyzed approaches to anomaly detection, time series forecasting, and log file analysis, as well as their limitations in the context of operational monitoring. Proposed a methodology for integrating external ML modules with Zabbix and a classification scheme for using ML models depending on the type of data and needed tasks. Performed a comparative analysis of ML approaches and formulated recommendations for their practical application, taking into account the requirements for achieving the target service level (SLO).