The statit output below is from a 12 disk SATA aggregate on a FAS960. The load is from a jetstress test that I let run only for a few minutes.
Disk Statistics (per second)
ut% is the percent of time the disk was busy.
xfers is the number of data-transfer commands issued per second.
xfers = ureads + writes + cpreads + greads + gwrites
chain is the average number of 4K blocks per command.
usecs is the average disk round-trip time per 4K block.
disk ut% xfers ureads--chain-usecs writes--chain-usecs cpreads-chain-usecs greads--chain-usecs gwrites-chain-usecs/aggsa/plex0/rg0:
3b.16 43 19.48 0.33 1.00 30786 17.86 30.41 1082 1.29 9.46 560 0.00 .... . 0.00 .... .
3b.17 43 19.43 0.33 1.00 36786 17.86 30.41 1099 1.24 8.60 485 0.00 .... . 0.00 .... .
2b.18 40 18.72 0.67 1.00 16893 17.43 30.88 1027 0.62 14.50 554 0.00 .... . 0.00 .... .
3b.19 40 18.00 0.00 .... . 17.17 31.33 1050 0.83 10.66 566 0.00 .... . 0.00 .... .
2b.20 40 17.69 0.00 .... . 17.07 31.47 1037 0.62 16.27 506 0.00 .... . 0.00 .... .
2b.21 42 17.67 0.00 .... . 17.10 31.46 1097 0.57 13.54 412 0.00 .... . 0.00 .... .
2b.22 41 17.76 0.02 1.00 16000 17.07 31.47 1115 0.67 13.11 474 0.00 .... . 0.00 .... .
3b.23 40 17.91 0.00 .... . 17.14 31.32 1045 0.76 13.53 545 0.00 .... . 0.00 .... .
2b.24 42 18.19 0.00 .... . 17.29 31.11 1160 0.91 8.76 1078 0.00 .... . 0.00 .... .
3b.25 41 17.95 0.02 1.00 9000 17.19 31.29 1077 0.74 9.77 614 0.00 .... . 0.00 .... .
2b.26 43 17.88 0.00 .... . 17.12 31.42 1167 0.76 12.72 747 0.00 .... . 0.00 .... .
3b.27 41 18.17 0.00 .... . 17.19 31.25 1103 0.98 11.29 585 0.00 .... . 0.00 .... .
My question is, how do you interpret this data correctly? For example, was 43% of 3b.16’s time busy issuing 17.86(ignoring reads and cpreads) write commands that wrote 30.41 4KB blocks (120KB’s) and took 1082(somethings?) to return.
Is an xfer the same thing as an I/O?
What is the measurement for usecs?
Thanks for the help.
- Carl
On 3/15/06 11:24 AM, "Blake Golliher" <thelastman@gmail.com> wrote:
I ususally gauge that I can get 40 to 50 iops / sec under a 10ms
responce time. So a little less then half of what a 10k RPM (fcal)
disk can do.
-Blake
On 3/15/06, Carl Howell <chowell@uwf.edu> wrote:
>
>
>
> Does anyone have any SATA performance data? I know that there are many
> variables when measuring disk i/o, so any information would be appreciated.
>
>
>
> Thanks,
>
>
>
> -Carl