An Analytical Computing Infrastructure for Monitoring Dynamic Networks Based on Knowledge Graphs

Abstract

Dynamic network monitoring systems are typically designed to solve a predefined number of tasks, new requirements lead to expensive development efforts and sometimes even require changes in the system architecture. Knowledge graphs are powerful and flexible tools for information integration and supported by a set of standardized vocabularies and languages (the “Sematic Web” toolset). In this work, we discuss the application of knowledge graphs to develop and analyze an analytical computing infrastructure for a dynamic network monitoring system. As a typical dynamic network, a multiservice telecommunication network is considered. The presented system combines static models of a telecommunication network and dynamic monitoring data and makes it possible to obtain complex analytical reports using SPARQL queries over the knowledge graph. Those reports are of crucial importance to network stakeholders for improving the network services and performance. First, we analyze problems solved by traditional monitoring systems, and identify the classes of problems such systems cannot solve. Then we propose an analytical monitoring system architecture based on knowledge graphs to address these classes of problems. We present the system structure and detailed descriptions of the ontological and mathematical models of the resulting knowledge graph. In order to test the architecture discussed, we create an example task of the analytical monitoring system and analyze system performance depending on the size of the knowledge graph. The results of the analysis are presented using a number of SPARQL queries.