Microsoft SQL Server database administrators usually perform pre-prepared
queries for diagnostics and monitoring, and consequently, from their results,
we infer the server status, which is time-consuming and error-prone in large
scale environments. We will design and implement an agentic system that
enables monitoring and optimization of the SQL server using natural lan
guage conversation. The system will be based on a multi-agent architecture,
where specialized agent nodes autonomously fulfil their roles. Communica
tion between agent nodes will be supported and coordinated by a software
orchestrator that does not know the semantics of the flow — the agent nodes
themselves dictate the execution sequence. The main contribution of the
thesis is an extensible architecture that improves and simplifies the classic
monitoring, optimization, and troubleshooting procedures in Microsoft SQL
Server systems.
|
