Umeå universitets logga

This study, conducted in collaboration with Ericsson Research, explores the potential of utilizing metric data for predictive analytics within IT operations. The primary objective is to address underutilized data by investigating its utility in forecasting future trends and behaviors. The research is driven by two key questions: to what extent can metric data inform predictive behaviors and the identification of specific metrics most valuable for predictive analysis? The study focuses on three main aims: evaluating the quality and predictive suitability of Zabbix-collected data, assessing the strength of correlations within the datasets using industry-standard analytical techniques, and developing an inference model based on identified metrics. Initial findings indicate that while the metric data holds significant potential for predictive analytics, it exhibits high individuality among hosts, requiring careful feature selection and temporal resolution analysis. This research lays the ground-work for future studies to utilze datasets at Ericsson Research.