Hi Alexander, Assuming the images would fit in the page cache of all your OSD nodes, you would see a massive performance increase as reads would be coming straight from ram. But otherwise no, reads are not balanced across replica's, only the primary one responds to reads. But don't forget a RBD is striped across all OSD's so you still get quite nice performance and as you will likely have high queue depths the IO's should be spread amongst the disks quite nicely. Nick From: ceph-users [mailto:ceph-users-bounces@xxxxxxxxxxxxxx] On Behalf Of alexander.dibbo@xxxxxxxxxx Sent: 13 March 2015 14:22 To: ceph-users@xxxxxxxxxxxxxx Subject: Many Reads of an object Hi, What would happen if an object in ceph was being read by many clients at the same time to the extent that the OSD holding the primary replica could not respond to the get requests? Would this trickle down to the next replica? Are the reads load balanced across the replicas in any way? We run a ceph cluster with OpenNebula to provide a private cloud. This is a replicated pool with connections through RBD. I realise that various caches and a number of other factors could come into play to change the likelihood of this happening in one way or the other. The scenario I am imagining is many virtual machines (hundreds, possibly thousands) are deployed and start to boot at the same time from a single master image and as such will be needing access to the same data at the same time I look forwards to hearing your thoughts on this. Thanks Alexander Dibbo _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |