Impact of Vector Search on RDBMS Performance

This technical report outlines the challenges and risks associated with integrating vector search capabilities into legacy relational database management systems (RDBMS). It discusses the operational issues that arise when combining Online Transaction Processing (OLTP) workloads with resource-intensive vector search tasks. The document details the architectural conflicts that occur, particularly highlighting the 'Noisy Neighbor' problem, where vector search operations compete for CPU and memory resources with critical transactional processes. The report presents a comprehensive analysis of the performance implications for major database platforms like PostgreSQL, Oracle, and SQL Server. It emphasizes the need for careful resource governance and tuning to mitigate the risks of performance degradation. Additionally, it provides strategic recommendations for maintaining system stability while leveraging new AI features. The report serves as a technical warning against the marketing narratives that promote integrated vector databases without addressing the inherent risks of resource contention.