No, it doesn't. FSAL_RGW effectively embeds a stripped down version of
the RGW daemon itself, and so communicates directly with the Ceph
cluster using the RADOS protocol, just like an RGW daemon does. It
likely won't be quite as fast as native S3 to an RGW daemon, due to
mismatches between the semantics of S3 and NFS, but it is currently
faster than FSAL_CEPH, I believe, due to the fact that libcephfs (the
access library used by FSAL_CEPH) is effectively single threaded due to
a single large lock.
On 8/26/21 7:26 PM, ethan.deson(a)oracle.com wrote:
Is the RGW_FSAL using the S3 API under the hood (aka sending S3 http
requests)? Someone on my team is conjecturing that RGW_FSAL will be markedly slower than
CEPH_FSAL on that basis.
Does anyone have a sense of the average performance characteristics of these two FSALs?
Support mailing list -- support(a)lists.nfs-ganesha.org
To unsubscribe send an email to support-leave(a)lists.nfs-ganesha.org