SeaweedFS is a distributed storage system designed around one problem most file systems handle badly: very large numbers of small files. It can serve them from disk in a single seek, and it exposes the result through an S3 API, a POSIX-like mount, WebDAV, and SFTP.

Architecture

Component	Role
Master	Tracks volumes and assigns file IDs. It is not in the read path
Volume servers	Store file data in large append-only volume files, with an in-memory index
Filer	Adds directories and file names on top, with pluggable metadata stores

Because volume servers pack many small files into big volume files and index them in memory, reading a file is one disk seek, and per-file metadata overhead is around 40 bytes. That is why SeaweedFS holds up where storing millions of thumbnails or log chunks as individual files would crush an ordinary filesystem.

Features

• S3-compatible object storage
• FUSE mount on Linux, macOS, and Windows
• WebDAV and SFTP
• Replication for hot data and erasure coding for warm data
• Tiering to cloud storage
• Apache Iceberg table support with a built-in REST catalog

Deployment

SeaweedFS ships as a single weed binary that can run every component, which makes a one-machine test easy:

weed server -dir=/srv/seaweedfs -s3

Production clusters run masters, volume servers, and filers separately, using Docker Compose, Kubernetes with Helm charts, or bare metal.

When to use it

SeaweedFS shines for large deployments and small-file-heavy workloads: media thumbnails, backups with many chunks, logs, and data lakes. For a single homelab box that just needs an S3 endpoint, Garage, RustFS, or MinIO are simpler to operate.

License

Apache License 2.0.

Links

• Repository