On Fri, Sep 9, 2016 at 10:33 AM, Alexandre DERUMIER <aderumier@xxxxxxxxx> wrote: > The main bottleneck with rbd currently, is cpu usage (limited to 1 iothread by disk) Yes, definitely a bottleneck -- but you can bypass the librbd IO dispatch thread by setting "rbd_non_blocking_aio = false" in your Ceph client config. Our long term librbd goal is to remove that IO dispatch thread once we can get a non-blocking IO path from end-to-end. -- Jason _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |