Hi, I've been trying to tweak and improve the performance of our ceph cluster. One of the operations that I can't seem to be able to improve much is the delete. From what I've gathered every time there is a delete it goes directly to the HDD, hitting its performance - the op may be recorded in the journal but I don't notice almost any impact. >From my tests (1M files with 512k) writing the data will take 2x as much as the delete operation - should there be a bigger difference? And whilst the delete operation is running all the remaining operations will be slower - it does impact the whole cluster performance in a significant way. Is there any way to improve the delete performance on the cluster? I'm using S3 to do all the tests, and the .rgw.bucket.index is already running from SSDs as is the journal. I'm running firefly 0.80.5. thanks, -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.ceph.com/pipermail/ceph-users-ceph.com/attachments/20140908/0eac3d40/attachment.htm>
 |