Overview

As the data scale in scenarios such as large-scale image retrieval, video search, image recognition, and recommendation increases, higher requirements are imposed on the latency and accuracy of high-dimensional space vector search. Based on the developed vector search engine and plug-ins of Elasticsearch, MRS integrates the vector search capability featuring high-performance, high-precision, low-cost, and multi-modal.

Compared with the brute-force search capability provided by Elasticsearch, MRS has more built-in vector index algorithms, accelerating the search and improving the efficiency. MRS provides functions such as distributed search, multi-copy search, error recovery, snapshot, and permission control compared with open-source vector search engines Faiss and NMSLIB. It is compatible with Elasticsearch and supports interconnection with Kibana, Garafa, and Logstash.