Hi,
Have just realised after running a bunch of iozone disk performance tests that iozone doesn't wait for 'fsync' to finish before declaring it's results finished. Noticed this when running the pts/iozone benchmark on the host RHEL system that serves a few VMs and seeing abnormally high write speeds on an ext4 filesystem. Had a quick look at the doco for iozone, and found that the '-e' argument forces it to include the fsync time in the calculation times. I have now modified the test-definition.xml for iozone to include the '-e' option as a default and am seeing much more realistic results for writes.
Michael, would you consider including this as a default for iozone in the next pts release, or if you won't do that, would you consider adding it as a selectable option for iozone?
Cheers
Have just realised after running a bunch of iozone disk performance tests that iozone doesn't wait for 'fsync' to finish before declaring it's results finished. Noticed this when running the pts/iozone benchmark on the host RHEL system that serves a few VMs and seeing abnormally high write speeds on an ext4 filesystem. Had a quick look at the doco for iozone, and found that the '-e' argument forces it to include the fsync time in the calculation times. I have now modified the test-definition.xml for iozone to include the '-e' option as a default and am seeing much more realistic results for writes.
Michael, would you consider including this as a default for iozone in the next pts release, or if you won't do that, would you consider adding it as a selectable option for iozone?
Cheers