| 1 |
hi |
| 2 |
|
| 3 |
Dale wrote, at 08/05/2012 04:45 PM: |
| 4 |
> Howdy, |
| 5 |
> |
| 6 |
> |
| 7 |
> I have heard of bonnie and friends. I also think dd could do some |
| 8 |
> testing too. Is there any other way to give this a good work and see if |
| 9 |
> it holds up? Oh, helpful hints with Bonnie would be great too. I have |
| 10 |
> never used it before. Maybe someone has some test that is really brutal. |
| 11 |
> |
| 12 |
|
| 13 |
some time ago i have played with bonnie++ to figure out my hard disk performance |
| 14 |
using different filesystems and io schedulers. the script invokes 3 bonnie |
| 15 |
instances; each instance runs its own set of tests: write/read/rewrite a 30gb |
| 16 |
file followed with different file operations on 48k small files spread over 32 |
| 17 |
sub-directories: |
| 18 |
|
| 19 |
#!/bin/bash |
| 20 |
# |
| 21 |
scratch=${1:-$(pwd)} |
| 22 |
# |
| 23 |
date |
| 24 |
ft=$(df -Pl $scratch|tail -1|awk '{print $6}') |
| 25 |
mnt=($(mount|grep "$ft ")) |
| 26 |
dev=$(basename $(readlink -fn ${mnt[0]})) |
| 27 |
sched=$(cat /sys/block/$dev/queue/scheduler|sed -e 's/.*\[//1' -e 's/\].*//1') |
| 28 |
log="$dev-${sched}-${mnt[4]}" |
| 29 |
echo $log |
| 30 |
rm -f ${log} ${log}.html |
| 31 |
/usr/sbin/bonnie++ -p 3 |
| 32 |
/usr/sbin/bonnie++ -qd $scratch -n48:128K:16K:32 -ys -s30g -mg1 >> ${log} & |
| 33 |
/usr/sbin/bonnie++ -qd $scratch -n48:128K:16K:32 -ys -s30g -mg2 >> ${log} & |
| 34 |
/usr/sbin/bonnie++ -qd $scratch -n48:123K:16K:32 -ys -s30g -mg3 >> ${log} & |
| 35 |
wait |
| 36 |
/usr/sbin/bonnie++ -p -1 |
| 37 |
bon_csv2html ${log} >> ${log}.html |
| 38 |
date |
| 39 |
|
| 40 |
this script worked around 30mins on a sata3 1tb drive. presuming your 3tb, you |
| 41 |
may adjust the file size and/or number of bonnie instances to fill up the disk |
| 42 |
space; then start the script and leave it running for a day. i guess this test |
| 43 |
would be brutal enough and on completion the disk might be considered good :) |
| 44 |
|
| 45 |
victor |
Archives