TITLE
PERFORMANCE ANALYSIS OF CHROMA, QDRANT, AND FAISS DATABASES
AUTHOR(S)
Emir Öztürk, Altan Mesut
ABSTRACT
The complexity and dimensions of deep learning models are increasing. Along with the growing complexity, vector databases have been proposed to store high-dimensional data required by the models. Vector databases aim to store high-dimensional vectors and perform similarity calculations on these vectors. In this study, the insertion and query performances of three different vector databases were measured on datasets of varying sizes, and the results were examined. The findings indicate that databases stored in main memory, such as Faiss, provide optimal performance without the need for an index in small-sized datasets and have fast response times. However, as the data size increases, the advantage diminishes with the increasing main memory requirement, and the use of Chroma, which provides index support for disk-stored data, becomes more suitable.
DOI
http://www.doi.org/10.70456/TBRN3643
DOWNLOAD
https://unitechsp.tugab.bg/images/2024/4-CST/s4_p72_v3.pdf
How to cite this article:
Emir Öztürk, Altan Mesut, PERFORMANCE ANALYSIS OF CHROMA, QDRANT, AND FAISS DATABASES, UNITECH – SELECTED PAPERS - 2024