SQLSIM: Consultas analíticas por similaridade em SGBD Relacionais

Abstract

With the exponential growth of multidimensional data, developing efficient approaches for similarity-based analytical queries has become essential. However, Relational Database Management Systems (RDBMS) do not natively support this type of query, often requiring external solutions that can impact efficiency and scalability. In this work, we propose SQLSIM, a strategy for executing similarity-based queries directly within RDBMS, eliminating the need for external processing and reducing impedance mismatch. The approach leverages user-defined functions (UDFs) in procedural SQL to implement similarity-based clustering operations across multiple dimensions. Additionally, we explore techniques to optimize the execution of these queries by utilizing indexing structures and advanced operators. The solution was implemented in PostgreSQL and evaluated through experiments that demonstrated its flexibility in processing large volumes of data. The results indicate that SQLSIM enables complex multidimensional analysis in an integrated manner, maintaining scalability and compatibility with standard SQL.