Problem

Having large objects in a replicated pool wastes capacity. Backup tools such as Veeam that write to S3 mix large numbers of small metadata/catalog objects with large backup-content objects in the same bucket. This hurts performance.

Solution

Use Ceph S3's support for dynamic placement and policy-driven data retention.

See

https://github.com/frednass/s3-dynamic-placement-and-archiving

See Also

• https://github.com/OSNEXUS/rgw-autotier - for information about attaching Lua scripts
• https://docs.ceph.com/en/reef/radosgw/lua-scripting/ - for Ceph-specific information about Lua scripting
• https://archive.fosdem.org/2026/schedule/event/WWQBMT-ceph-s3-dynamic-placement-and-optimized-retention/ - Frederic Nass's talk on this subject