Introduction
BlixtFS turns a cloud bucket into a shared filesystem that any application can use, without copying the data or changing its format.
What BlixtFS is
Object storage is the cheapest and most durable place to keep large amounts of data. Most software, though, expects a filesystem: it opens paths, lists directories, appends to files and renames them. Rewriting that software to call an object-storage API is expensive, and the usual workaround is to copy data out of the bucket onto disks, which costs time and money.
BlixtFS bridges that gap. It runs next to your bucket and serves it as a POSIX-style filesystem over standard protocols supported by your operating system:
- NFS (v3 or v4) for Linux, MacOS and Unix servers.
- SMB for Windows and MacOS desktops
A client mounts BlixtFS like any other network share. Nothing needs to be installed on the client.
Other file protocol options are available:
- Kubernetes CSI for Kubernetes.
- MPI I/O for HPC workloads.
- SFTP for file transfers.
What makes it different
Your data stays yours. Every file is stored as one ordinary object, under the same name you see in the filesystem. There is no proprietary block format and no BlixtFS-only encoding. You can keep reading and writing the bucket directly with your provider’s own tools, and you can stop using BlixtFS at any time without migrating anything.
It runs in your cloud account. BlixtFS is software you deploy, not a hosted service. Data goes straight between your servers and your bucket, and cloud credentials never leave the servers you give them to.
It is fast. Blixt’s custom metadata index answers directory listings and file lookups without calling the object store. File data is read efficiently from the object stores and cached locally. Writes are persistend locally and uploaded in the background. See Architecture for the details.
It is correct.
It keeps up with external changes. When another program writes to the bucket directly, BlixtFS is updated through the cloud provider’s change notifications and updates the filesystem. See Change notifications.
It works with any major object store. One deployment can serve buckets from several providers side by side. See Cloud providers.
Who it is for
BlixtFS suits workloads that want object-storage economics but need file semantics. For example:
- AI and machine-learning training, where many nodes read large datasets
- Media production, where editors work on shared footage from desktops
- Office file shares, replacing an on-premises file server with a bucket
- Multi-cloud data access, presenting buckets from several providers under one mount
Use cases describes each in more detail.
Where to go next
- Key concepts: the vocabulary used throughout these docs
- Editions: what the Standard, Enterprise and High Performance editions include
- Quickstart: mount a bucket with Docker in a few minutes