Ceph is a distributed storage platform that provides object, block, and file storage on a shared cluster of commodity hardware.
tigrisdata/tigrisfs
A FUSE file system that mounts S3-compatible object storage buckets as local directories, aimed at AI workloads.
Open-source alternative to:Amazon S3 Mountpoint
- Stars
- 245
- Forks
- 13
- Open issues
- 8
- Last commit
- 24 days ago
What is it?
TigrisFS mounts an S3 or S3-compatible object store bucket as a local file system through FUSE, so programs can read and write bucket contents with ordinary file paths. It is written in Go and is a fork of GeeseFS, which was itself derived from Goofys. The project focuses on the performance problems that S3-backed FUSE file systems typically have with small files and metadata operations, using parallelism and asynchronous requests to address them.
When mounted against the Tigris backend, TigrisFS supports POSIX permissions, special files and symbolic links, auto-preloads the content of small files when a directory is listed, and can prefetch directory data to the region on list. The project also removes a bundled, outdated AWS SDK, upgrades dependencies, and runs the race detector and linters in CI.
Installation is available through a one-line install script, prebuilt DEB and RPM packages with systemd units for mounting buckets, or a binary tarball for macOS ARM64. Credentials can come from standard AWS credential files or environment variables.
Key features
- 01Mounts S3-compatible buckets as local file system paths via FUSE.
- 02Uses parallel and asynchronous operations to speed up small-file and metadata access.
- 03Supports POSIX permissions, special files and symbolic links on the Tigris backend.
- 04Auto-preloads small file contents when a directory is listed.
- 05Reads AWS credentials from standard config files or environment variables.
- 06Ships as DEB, RPM and binary packages with systemd mount units.
Use cases
- Giving AI training or inference code direct file access to datasets stored in object storage.
- Mounting a cloud bucket on a Linux server to work with its files using standard tools.
- Running distributed cluster jobs that read shared data from an S3-compatible bucket.
Quick start
- 1
Run the official one-line install script from the README
- 2
Configure S3 credentials via AWS config files or environment variables
- 3
Mount the bucket as the current user
/usr/local/bin/tigrisfs <bucket> $HOME/mnt/tigrisfs/<bucket>
Things to know
- Prebuilt binaries in the README cover only Linux amd64 packages and macOS ARM64.
- Mounting requires S3-compatible credentials and a FUSE-capable system.
Topics
FAQ
- What is tigrisfs?
- A FUSE file system that mounts S3-compatible object storage buckets as local directories, aimed at AI workloads. TigrisFS mounts an S3 or S3-compatible object store bucket as a local file system through FUSE, so programs can read and write bucket contents with ordinary file paths. It is written in Go and is a fork of GeeseFS, which was itself derived from Goofys. The project focuses on the performance problems that S3-backed FUSE file systems typically have with small files and metadata operations, using parallelism and asynchronous requests to address them.
- Is tigrisfs free to use? What license does it have?
- tigrisfs is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
- Is tigrisfs still maintained?
- Actively maintained. The last commit to tigrisfs was on Sep 29, 2026.
- What can I use tigrisfs for?
- Giving AI training or inference code direct file access to datasets stored in object storage. Mounting a cloud bucket on a Linux server to work with its files using standard tools. Running distributed cluster jobs that read shared data from an S3-compatible bucket.
- How do I install tigrisfs?
- Run the official one-line install script from the README Configure S3 credentials via AWS config files or environment variables Mount the bucket as the current user: /usr/local/bin/tigrisfs <bucket> $HOME/mnt/tigrisfs/<bucket>
- Where can I get tigrisfs?
- The source code and installation instructions are on GitHub: https://github.com/tigrisdata/tigrisfs
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A distributed storage system that serves S3 object storage, POSIX file systems, and Iceberg/Lance tables from one binary, built for billions of small files.
A Go tool that mounts multiple cloud drives and S3/WebDAV-style storage as one local folder, with rclone-compatible end-to-end encryption.
A stateless, Kafka-compatible streaming system that stores data on S3 or S3-compatible object storage instead of local disks.
S3-compatible distributed object storage for small-to-medium self-hosted clusters spread across multiple physical locations.
Reference Go server for the tus protocol, enabling resumable HTTP file uploads to local disk, Google Cloud Storage or S3.