Published on march 25, 2023
AWS announced Mountpoint for Amazon S3, an open-source file client that provides high-throughput access to Amazon S3. As of now, the local mount point provides high single-instance transfer rates and is intended primarily for data lake applications.
A mountpoint for Amazon S3 translates local file system API calls into S3 object API calls such as GET and LIST. It supports both random and sequential reading of files as well as listing of files and directories. The alpha release does not support writes (PUTs), and future versions of the client will only support sequential writes to new objects.
James Bornholt, scholar at AWS and assistant professor at the University of Texas, Devabrat Kumar, senior product manager at AWS, and Andy Warfield, distinguished engineer at AWS, acknowledge that the client is not a general-purpose networked file system, and some file operations are restricted.
The open-source client does not emulate operations such as directory renaming that would require numerous S3 API calls or POSIX file system features that are not supported by S3.
As with Goofys and s3fs, Mountpoint for S3 is not the first client to present S3 as a filesystem.
In order to validate the file system semantics, the new client is built in Rust on the Common Runtime (CRT) used by most AWS SDKs.
Mountpoint is released under the Apache License 2.0, but is not yet ready for production use. On GitHub, you can find the initial alpha release and the public roadmap.
