IOzone is reporting that my consumer NVMe SSD drive's performance was:
IF the units-of-measure ... was wrong ... as in, not KB but Kb, or Bytes, ... even then, the numbers are nutty.
TO SEE THIS SCREEN SHOT UN-EDITED ... DOWNLOAD THE FILE: IOzone condensed.png
The results of this highly-respected & oft-recommended utility are confusing & contradicted by other benchmarking tools.
If I'm correct, I hope it helps others (or alerts the mfr to fix the issue, If inaccurate) ... If I'm wrong, please explain what I missed.
Suggestions (or corrections) anyone...?
XLS attached files includes a screenshot of the CLI output ... but removed (in the the 'Random RW + CLI file) all but random RW info...
XLS attached named "output - IOZone.xls" is the original (unedited) output -- which I cross referenced to the CLI data to confirm it's the same.
- Random 4 KB Reads @ 800,000 IOPS ..!
- Reads at 16 GB/s ...
IF the units-of-measure ... was wrong ... as in, not KB but Kb, or Bytes, ... even then, the numbers are nutty.
- Recalculating the units (instead of KB) as bytes, then, instead of 16 GB/s it gets 16 MB/s !?
- Recalculating the units (instead of KB) as Kbits, then, instead of 16 GB/s it gets 2 GB/s - of mixed 16KB RANDOM reads..??
TO SEE THIS SCREEN SHOT UN-EDITED ... DOWNLOAD THE FILE: IOzone condensed.png
The results of this highly-respected & oft-recommended utility are confusing & contradicted by other benchmarking tools.
If I'm correct, I hope it helps others (or alerts the mfr to fix the issue, If inaccurate) ... If I'm wrong, please explain what I missed.
Suggestions (or corrections) anyone...?
XLS attached files includes a screenshot of the CLI output ... but removed (in the the 'Random RW + CLI file) all but random RW info...
XLS attached named "output - IOZone.xls" is the original (unedited) output -- which I cross referenced to the CLI data to confirm it's the same.
Attachments
Last edited: