What do you use as an interconnect between your osds, and your clients?
On Mar 20, 2016 11:39 AM, "Mike Almateia" <mike.almateia@xxxxxxxxx> wrote:
18-Mar-16 21:15, Schlacta, Christ пишет:
Insofar as I've been able to tell, both BTRFS and ZFS provide similar
capabilities back to CEPH, and both are sufficiently stable for the
basic CEPH use case (Single disk -> single mount point), so the
question becomes this: Which actually provides better performance?
Which is the more highly optimized single write path for ceph? Does
anybody have a handful of side-by-side benchmarks? I'm more
interested in higher IOPS, since you can always scale-out throughput,
but throughput is also important.
By now, we have a small cluster (5 node * 12 HDD).
Btrfs on the disks - one disk, one OSD.
Also we use EC 3+2.
Cluster build for video record from street's cams.
Our tests with 4Mb bs/99% seq. write show a good perfomance:
* around 500-550 Mb/s with BTRFS vs 120-140 Mb/s Disk+Journal on same disk.
We use Centos 7.2.
--
Mike.
_______________________________________________
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
