Sinema: Semantics-driven Intelligent Network Management using AI assistance

The growing complexity of network infrastructures has resulted in the generation of vast amounts of network management data, presenting significant challenges in efficient and effective network reliability and security management. Traditional log analysis techniques are increasingly insufficient to handle this complexity, necessitating the development of more advanced solutions. In this paper, we propose Sinema, a novel approach that integrates a knowledge graph with an AI-agent for network management data analysis. The system transforms raw Syslog data into a structured graph format, enabling sophisticated querying and analysis. Sinema’s performance is evaluated across two key dimensions: retrieval and analysis. Regarding the efficiency of data retrieval, quantitative performance metrics are used to assess query accuracy and completeness. As for the effectiveness of analysis, a qualitative measure is employed, focusing on the diversity and correctness of the analysis questions the system can answer. The evaluation results demonstrate Sinema’s robustness in handling complex queries and its capability to provide accurate insights, contributing a scalable and efficient framework for network management data analysis that enhances proactive network management and threat detection.