Using Ceph like S3 could be a bit tricky if all of that 10pb is very small files. Redhat did an interesting series of blog posts about Ceph and getting it to 1

Using Ceph like S3 could be a bit tricky if all of that 10pb is very small files. Redhat did an interesting series of blog posts about Ceph and getting it to 1 billion objects https://www.redhat.com/en/blog/scaling-ceph-billion-objects-and-beyond https://www.redhat.com/en/blog/scaling-ceph-billion-objects-...