# A Comprehensive Survey on Vector Database: Storage and Retrieval Technique, Challenge

**Authors:** Yikun Han, Chunjiang Liu, Pengfei Wang  
**arXiv:** [2310.11703](https://arxiv.org/abs/2310.11703)  
**Date:** 2023-10-18  
**Summary page:** https://www.summarizepaper.com/en/paper/2310.11703v1/

## Key points

- Vector databases are designed to store high-dimensional data that traditional database management systems struggle to handle.
- The focus on the approximate nearest neighbor search (ANNS) problem has been a longstanding area of study with numerous algorithmic articles available in the literature.
- The authors categorize studies based on their approach to solving the ANNS problem, including hash-based, tree-based, graph-based, and quantization-based methods.
- Organizing algorithms within a framework helps readers understand diverse strategies for addressing challenges in high-dimensional data storage and retrieval.
- The article highlights existing challenges faced by vector databases and explores potential solutions to enhance their performance.
- One intriguing aspect discussed is integrating vector databases with large language models for new possibilities in data processing and analysis.
- This survey serves as a valuable resource for researchers and practitioners seeking insights into cutting-edge techniques for managing high-dimensional data effectively.

## Summary

The article "A Comprehensive Survey on Vector Database: Storage and Retrieval Technique, Challenge" by Yikun Han, Chunjiang Liu, and Pengfei Wang delves into the realm of vector databases. These databases are designed to store high-dimensional data that traditional database management systems struggle to handle. Despite the scarcity of literature discussing existing or innovative vector database architectures, the focus on the approximate nearest neighbor search (ANNS) problem has been a longstanding area of study with numerous algorithmic articles available in the literature. The authors aim to provide a thorough review of relevant algorithms in this burgeoning research field. They categorize studies based on their approach to solving the ANNS problem. These approaches include hash-based, tree-based, graph-based, and quantization-based methods. By organizing these algorithms within a framework, readers can gain a better understanding of the diverse strategies employed in addressing the challenges posed by high-dimensional data storage and retrieval. Furthermore, the article highlights the existing challenges faced by vector databases and explores potential solutions to enhance their performance. One intriguing aspect discussed is how vector databases can be integrated with large language models to unlock new possibilities for data processing and analysis. Overall, this comprehensive survey serves as a valuable resource for researchers and practitioners seeking insights into cutting-edge techniques for managing high-dimensional data effectively. It sheds light on the evolving landscape of vector databases and offers a roadmap for future advancements in this dynamic field.