On 7/2/15, 9:21 AM, "Nate Curry" <cu...@mosaicatm.com<mailto:cu...@mosaicatm.com>> wrote:
Are you using the 4TB disks for the journal? Nate - yes, at the moment the Journal is on 4 TB 7200 rpm disks as well as the OSDS. It's what I've got for hardware ... sitting around in 60 servers that I could grab. I realize it's less than idea - but ... beggars ... I've structured my OSDs with 1 spinning disk as Journal to 5 OSDs, and a second spinning disk as Journal to 5 other OSDs in each node. That was my basic plan. I could as easily go with 3 Journals each with 3 OSDS per Journall disk. I'd be interested to see how that changes performance across the cluster ... Future (our "Gen3") hardware platforms have SSDs in them ... We're looking to get a good understanding of what cluster with spinning disks on truly commodity hardware can do - and then compare to faster sexier hardware. We do have a lot of use cases for slower object services, where super high IOPS and performance isn't critical; that’s our starting point for testing. ~~shane
_______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
