I have been running some tests on LVM on 2.4.9-6 kernel from redhat, 1.0.1rc4. I have created a test volume, 64k stripe size, 26 stripes, 9Gig size, 8Gig RAM. Here is the odd part. I am using iozone to benchmark the disks. I start a 9Gig file test. The write gets about 56MB/sec, the re-write gets 68MB/sec, and the read get a whopping 13MB/sec. This seems completely backward to me. I'm not subscribed to lvm-devel, but a similar question has come up there under the topic, [lvm-devel] stripe performance (read_ahead vs. max_readahead) omn the 15th of October. Can anyone shed any light on this? The PV's are mirrored Fiber Channel disks. All separate disks. I've tried different FC drivers with the same results. We're about to run some Oracle tests, I was just checking the relative speeds. Is there any more information I can provide to help speed this up? Thanks Robert
