Not at the moment. We had some discussions about "blind" buckets, it's definitely on our mind, but we're not there yet. Yehuda On Sun, Mar 17, 2013 at 3:09 PM, Rustam Aliyev <rustam.lists@xxxxxxx> wrote: > Thanks for detailed explanation. > > Is there any way to disable bucket indexes? We already store index in our > Cassandra cluster and need RADOS only to store objects. We don't plan to do > any listing operations, only PUT and GET. > > > > On 17/03/2013 16:24, Gregory Farnum wrote: >> >> RADOS doesn't store a list of objects. The RADOS Gateway uses a separate >> data format on top of objects stored in RADOS, and it keeps a per-user list >> of buckets and a per-bucket index of objects as "omap" objects in the OSDs >> (which ultimately end up in a leveldb store). A bucket index is currently a >> single object (stored on one OSD), so that is a performance (and, much >> later, storage) bottleneck that you can run into with extremely large >> buckets that see enough traffic. We don't have a precise number (it depends >> in large part on how powerful your OSDs are) but it's somewhere in the many >> millions of objects. >> -Greg >> Software Engineer #42 @ http://inktank.com | http://ceph.com >> >> >> On Monday, March 11, 2013 at 4:28 PM, Rustam Aliyev wrote: >> >>> Thanks Sam, >>> >>> That's great. I'm trying to understand a bit of RADOS internals and I >>> went through architecture wiki. Yet unclear about some points. >>> >>> Where does RADOS store the list of objects (object metadata)? According >>> to RADOSGW docs S3 bucket listing is available, so it must be stored >>> somewhere. >>> >>> Thanks, >>> Rustam. >>> >>> On 11/03/2013 20:28, Sam Lang wrote: >>>> >>>> On Fri, Mar 8, 2013 at 9:57 PM, Rustam Aliyev <rustam.lists@xxxxxxx >>>> (mailto:rustam.lists@xxxxxxx)> wrote: >>>>> >>>>> Hi, >>>>> >>>>> We need to store ~500M of small files (<1MB) and we were looking to >>>>> RadosGW >>>>> solution. We expect about 20 ops/sec (read+write). I'm trying to >>>>> understand >>>>> how Monitoring nodes store Crush maps and what are the limitations. >>>>> >>>>> For instance, is there any recommended max number of objects per >>>>> Monitoring >>>>> node? Does adding more monitoring nodes will help to scale number of >>>>> small >>>>> files and iops (assuming that OSDs not a bottleneck)? >>>> >>>> >>>> No, the number of objects is unrelated to the monitors, so 3 monitors >>>> should suffice in your case. >>>> -sam >>>> >>>>> Many thanks, >>>>> Rustam. >>>>> _______________________________________________ >>>>> ceph-users mailing list >>>>> ceph-users@xxxxxxxxxxxxxx (mailto:ceph-users@xxxxxxxxxxxxxx) >>>>> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com >>> >>> >>> >>> _______________________________________________ >>> ceph-users mailing list >>> ceph-users@xxxxxxxxxxxxxx (mailto:ceph-users@xxxxxxxxxxxxxx) >>> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com >> >> >> > > _______________________________________________ > ceph-users mailing list > ceph-users@xxxxxxxxxxxxxx > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |