On 15/12/14 17:44, ceph.com@xxxxxxxxxxxxx wrote:
I have the following setup: Node1 = 8 x SSD Node2 = 6 x SATA Node3 = 6 x SATA
Having 1 node different from the rest is not going to help...you will probably get better results if you sprinkle the SSD through all 3 nodes and use SATA for osd data and the SSD for osd journal.
Client1 All Cisco UCS running RHEL6.5 + kernel 3.18.0 + ceph 0.88. A "dd bs=4k oflag=direct" test directly on a OSD disk shows me: Node1 = 60MB/s Node2 = 30MB/s Node2 = 30MB/s
Hmmm - your SSD are slow for direct writes (15K IOPS if my maths is right - what make and model are they)? For that matter your SATA seem a bit pretty slow too (what make and model are they)?
And as Christian has mentioned, ceph small block size IO performance has been discussed at length previously, so it is worth searching the archives to understand the state of things and see that there has been *some* progress with improving this issue.
Cheers Mark _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
