juicefs VS s3-benchmark

Support for Kerberos keytab files

JuiceFS under the Apache License 2.0, is a high-performance POSIX file system optimized for cloud-native environments. It stores data in Object Storage (e.g., Amazon S3) and metadata in databases like Redis, MySQL, or TiKV. JuiceFS integrates massive cloud storage with big data, machine learning, and AI applications efficiently, akin to local storage. It features full POSIX and Hadoop compatibility, S3 interface, Kubernetes support, and shared file storage for numerous clients. Some cool features are - strong consistency, scalable performance, data encryption, global file locks, and compression with LZ4 or Zstandard.

To improve the handling process of capacity overrun, the JuiceFS client supports deletion operations in the case of Ceph cluster fullness (see related code changes in JuiceFS Community Edition). Therefore, for newer client versions, there is no need to use set-full-ratio for temporary adjustments.

If you have any questions or would like to learn more, feel free to join discussions about JuiceFS on GitHub and the JuiceFS community on Slack.

The architecture image shows GCS and others, so I suspect it does.

https://github.com/juicedata/juicefs#architecture

See also: JuiceFS: https://juicefs.com/

Adds a DBMS or key-value store for metadata, making the filesystem much faster (POSIX, small overwrites don't have to replace a full object in the GCS/S3 backend).

Almost certainly a better solution if you want to turn your object storage into a mountable filesystem, with the (big) caveat that you can't access the files directly in the bucket (they are not stored transparently).

JuiceFS can do the same thing as s3fs-fuse, but better. Because it supports robust data consistency and caching policies to improve performance.

There's another benchmark somewhere showing S3 can max out a 100Gbps instance.

https://github.com/dvassallo/s3-benchmark

Another potential issue is ListBucket rate limiting. If you have lots of small objects, you'll spend most of the time waiting to discover the names than transferring data

TTFB in S3 is 20-30ms around the 50th percentile. it can go much higher at p99 [1]. In any case, rotational latency for HDD drives is an order of magnitude lower (typically 2-5ms for a seek operation).

S3 is great for higher throughput workloads where TTFB is amortized across larger downloads (this is why it's very common to use S3 as a "data lake" where larger columnar files are stored, usually at the order of hundreds of MiB).

I think it's an interesting project but perhaps explaining the use cases where this solution is beneficial would go a long way here.

[1] https://github.com/dvassallo/s3-benchmark