Whitepaper
Milvus Performance Report
June 2024
Milvus is an open-source vector database built to power vector similarity search and various GenAI use cases, such as Retrieval Augmented Generation (RAG). In this benchmark report, we showcase Milvus's performance through comprehensive metrics like throughput, latency, and recall rate, utilizing the open-source VectorDBBench across four real-world datasets from OpenAI and Cohere.
What You’ll Learn?
Are you considering Milvus for your applications and curious about its real-world performance or how it handles your specific datasets? Do you need guidance on evaluating alternative vector databases or comparing different vector databases based on performance metrics? In this guide, you’ll find your answers.
We cover the following key topics in this paper:
- Milvus Performance with Real-World Datasets: Gain insights into Milvus's capabilities using actual datasets from industry leaders.
- Evaluating Milvus with Your Own Datasets: Learn how to assess Milvus's performance with your specific data.
- Comparing Vector Databases: Understand how to use VectorDBBench, an open-source benchmarking tool, to evaluate and compare the performance of different vector databases.
- Factors Influencing Performance: Discover the key factors that affect vector database performance.
A Snapshot of Our Benchmark Approach
Our benchmarking tests are designed to provide a consistent and up-to-date performance evaluation of Milvus, reflecting its latest advancements. We periodically revisit and re-run these benchmarks to ensure the data remains current and relevant. All benchmark code is openly accessible on GitHub, and we encourage the developer community to engage, contribute, and improve through issues, comments, or pull requests.
Real-Time Progress Tracking
Curious about our ongoing progress? Check out our real-time leaderboard powered by VectorDBBench, an open-source tool dedicated to delivering unbiased benchmark results for popular vector databases and cloud services. VectorDBBench is user-friendly, simplifying the comparison of vector database performance and cost-effectiveness, making it accessible even to non-professionals.
Share
Get the Whitepaper