Milvus
Scalable open-source vector database for semantic search and RAG at large data volumes.
Quick answer
Milvus is a scalable open-source vector database for semantic search and RAG at large data volumes, listed under Data labeling & vector databases. A typical small-business use: vector store behind a company-wide RAG search over millions of document chunks. Official hardware: minimum 8 GB RAM (recommended 4 vCPU · 16 GB RAM). That fits a Hetzner Cloud CPX32 server at about €35.99/month (server only, excl. VAT; €69.99/month for the recommended spec). License: Apache-2.0, which allows internal business use.
Last updated · facts verified against official docs on · sources
Activity data comes from GitHub and was last refreshed Oct 22, 2026.
What is Milvus?
Milvus stores and searches embedding vectors with filtering, hybrid (dense + sparse) and multi-vector search, and scales from a single "standalone" server to a distributed cluster. Many RAG and agent frameworks support it as a vector store. The standalone Docker Compose setup runs Milvus with etcd for metadata and MinIO for object storage, plus a built-in web UI.
Where it fits in a 10–200 person business
- Vector store behind a company-wide RAG search over millions of document chunks
- Product, image or support-ticket similarity search on your own infrastructure
- Back-end for a knowledge-base app that has outgrown an embedded vector store
These use cases are our editorial suggestions, not claims made by the project.
Can a business use Milvus? (license)
| License | Apache-2.0 |
|---|---|
| Type | Permissive: MIT / Apache-2.0: use, modify and sell with few obligations. |
| Internal business use | Allowed under the license text |
| What to watch | Permissive. Internal and commercial use allowed; keep license and notices. |
This is a plain-English summary of the license text, not legal advice. Third-party models you run with it have their own licenses.
Data & compliance: does Milvus phone home?
| Telemetry (usage data sent to the vendor) | Not documented |
|---|---|
| Other calls home by default (updates, marketplaces, default cloud services) | Not documented |
| Single sign-on (OIDC / SAML / LDAP) | Not documented |
| Audit log | Not documented |
| Role-based access control (RBAC) | Yes, free The project lists mandatory user authentication, TLS encryption and role-based access control (RBAC) among its security features. |
| Data residency / GDPR notes from the project | Not documented |
Checked against the project's own docs and repository on . "Not documented" means we could not find the answer in official sources; it does not mean the feature is missing or that no data is sent. Third-party services you connect (model APIs, vector databases, search engines) have their own data policies. Compare all apps.
What hardware does Milvus need?
| Minimum | 8 GB RAM |
|---|---|
| Recommended | 4 vCPU · 16 GB RAM |
Standalone figures (cluster needs 32 GB minimum, 128 GB recommended). CPU must support SSE4.2, AVX, AVX2 or AVX-512. SATA SSD required, NVMe recommended. RAM and disk scale with data volume.
How much does it cost to host Milvus per month?
| Minimum spec | 8 GB RAM → CPX32 (4 vCPU, 8 GB, 160 GB disk) €35.99/mo |
|---|---|
| Recommended spec | 4 vCPU · 16 GB RAM → CPX42 (8 vCPU, 16 GB, 320 GB disk) €69.99/mo |
How we got this: we took the cheapest Hetzner Cloud Regular Performance (shared AMD vCPU) plan with at least the RAM and vCPU the project lists, using prices read from www.hetzner.com/cloud/regular-performance on 2026-10-22. "From" monthly price cap shown by Hetzner, excluding VAT. Prices vary by location. Covers the server only; it excludes LLM API usage, backups, extra storage, GPUs and your time. Disk is not matched; add a volume if you need more than the plan includes.
How to deploy Milvus
Method: Docker Compose (standalone)
# use the current release tag from https://github.com/milvus-io/milvus/releases
wget https://github.com/milvus-io/milvus/releases/download/v3.0.3/milvus-standalone-docker-compose.yaml -O docker-compose.yaml
sudo docker compose up -d
# gRPC/REST on :19530, web UI on http://127.0.0.1:9091/webui/From the official instructions: github.com/milvus-io/milvus-docs/blob/v3.0.x/site/en/getstarted/run-milvus-docker/install_standalone-docker-compose.md. Always check the source for the current version before you run it.
Known gotchas when self-hosting Milvus
- etcd is very sensitive to disk latency. Use local NVMe SSDs; slow disks can cause frequent leader elections that degrade the service. The docs give a fio test (over 500 IOPS, under 10 ms p99 fsync). source
- The bundled MinIO container runs with default credentials. Change them and keep ports 9000/9001 off the public network. source
- Upgrading a 2.5.x instance that uses RocksMQ to 2.6.x or later requires pinning mq.type rocksmq in user.yaml first, because changing the message queue during an upgrade is not supported. source
- On macOS, give the Docker Desktop VM at least 2 vCPUs and 8 GB of memory, otherwise installation might fail. source
Alternatives in Data labeling & vector databases
- Qdrant Fast vector database for semantic search and RAG back-ends.
- Label Studio Data labeling tool for text, images, audio and video, used to build and evaluate AI models.
Milvus FAQ
How much RAM does Milvus need?
The official minimum is 8 GB RAM. The recommended spec is 4 vCPU · 16 GB RAM. Standalone figures (cluster needs 32 GB minimum, 128 GB recommended). CPU must support SSE4.2, AVX, AVX2 or AVX-512. SATA SSD required, NVMe recommended. RAM and disk scale with data volume.
How much does it cost to self-host Milvus?
The software is free to download. Based on the official minimum 8 GB RAM (recommended 4 vCPU · 16 GB RAM): That fits a Hetzner Cloud CPX32 server at about €35.99/month (server only, excl. VAT; €69.99/month for the recommended spec). That covers the server only, not LLM API usage, backups, extra storage or your time.
Is Milvus free for commercial use?
Milvus is licensed under Apache-2.0 (Permissive). Internal business use is allowed under the license text. Permissive. Internal and commercial use allowed; keep license and notices. This is a summary, not legal advice.
What are common problems when self-hosting Milvus?
etcd is very sensitive to disk latency. Use local NVMe SSDs; slow disks can cause frequent leader elections that degrade the service. The docs give a fio test (over 500 IOPS, under 10 ms p99 fsync). The bundled MinIO container runs with default credentials. Change them and keep ports 9000/9001 off the public network. Upgrading a 2.5.x instance that uses RocksMQ to 2.6.x or later requires pinning mq.type rocksmq in user.yaml first, because changing the message queue during an upgrade is not supported. On macOS, give the Docker Desktop VM at least 2 vCPUs and 8 GB of memory, otherwise installation might fail.
Is Milvus still maintained?
Yes, it is actively developed. Last commit: Oct 22, 2026. Latest release: v3.0.3 (Oct 14, 2026). Checked Oct 22, 2026.
How do I install Milvus?
The official method is Docker Compose (standalone). The exact commands are in the "How to deploy" section above and in the official instructions: https://github.com/milvus-io/milvus-docs/blob/v3.0.x/site/en/getstarted/run-milvus-docker/install_standalone-docker-compose.md
What are self-hosted alternatives to Milvus?
Other apps in Data labeling & vector databases: Qdrant, Label Studio.
Sources
- GitHub repository: github.com/milvus-io/milvus
- Official documentation: milvus.io/docs/install_standalone-docker-compose.md
- License text: github.com/milvus-io/milvus/blob/master/LICENSE
- Hardware requirements, Gotcha 1, Gotcha 4: github.com/milvus-io/milvus-docs/blob/v3.0.x/site/en/getstarted/run-milvus-docker/prerequisite-docker.md
- Deploy instructions, Gotcha 2, Gotcha 3: github.com/milvus-io/milvus-docs/blob/v3.0.x/site/en/getstarted/run-milvus-docker/install_standalone-docker-compose.md
- Hetzner Cloud pricing (cost estimate): www.hetzner.com/cloud/regular-performance
- GitHub stars, commits and releases: GitHub REST API, fetched
Facts on this page were checked against the project's own repository and docs on 2026-10-22. Spotted something out of date? Tell us.