Multikriteriální analýza a predikce incidentů v síti

Abstract

Network technologies today can handle very high transfer speeds. At this monent, it is common to have 10Gbps even 100Gbps links. It is still more difficult to do network flow analysis because of high speed and needs of more powerful hardware. Todays methods are based mainly to analyze data flows and patterns. It is quite impossible to find abnormalities on high speed link because small amount of change in flows can mean big excesses in incidents. Thesis will focus on multi criteria analysis of data to be able to predict incidents. From this reason, monitoring system seems to be good core for it. Data from monitoring systems are available for bigger community than only owners of hardware probes. It is enough to use right mathematical methods to analyze data and get useful informations. Some interesting abnormalities in network were found even using simple correlations of data. Next stage was to prepare and preprocess data to classify servers and hosts by their behavior. It is possible to say that deeper analysis is possible due to Zabbix monitoring system and its features like Open-Source core, documented API and SQL backend for data. Benefit of this thesis is new approach to analysis containing algoritms which allow to identify significant items in monitoring system.