Building a NoSQL-Based Search Engine: A Scalable Approach to Information Retrieval

Abstract

This thesis presents Aurion, a search engine architecture built on NoSQL databases that addresses the scalability and performance limitations of some traditional relational database-based search systems. The system implements multiple ranking algorithms including TF-IDF, BM25, and PageRank, integrated with a distributed NoSQL infrastructure comprising MongoDB, ArangoDB, Cassandra, and Redis. Experimental results demonstrate that the NoSQL-based architecture achieves sub-second query response times while handling millions of documents, with horizontal scalability properties that some traditional systems cannot match. The implementation includes a modern web interface built with React and TypeScript, a FastAPI backend, and comprehensive analytics capabilities. Performance benchmarks show average query response times of 247ms across indexed documents, with linear scalability and production-ready deployment on Google Cloud Platform. This work contributes to the field of information retrieval by demonstrating practical implementations of polyglot persistence for large-scale search applications and provides comparative analysis of different ranking algorithms in distributed environments.