| 1 |
On Sun, Jan 23, 2011 at 4:02 AM, Dale <rdalek1967@×××××.com> wrote: |
| 2 |
> Mark Knecht wrote: |
| 3 |
>> |
| 4 |
>> On Sat, Jan 22, 2011 at 1:15 AM, Dale<rdalek1967@×××××.com> wrote: |
| 5 |
>> <SNIP> |
| 6 |
>> |
| 7 |
>>>>> |
| 8 |
>>>>> Other than copying a file and using time to measure how long it takes, |
| 9 |
>>>>> what is the best test of a hard drive's speed? |
| 10 |
>>>>> |
| 11 |
>>>> |
| 12 |
>>>> By running a benchmark tool that does exactly this. IOzone is a nice |
| 13 |
>>>> one: |
| 14 |
>>>> |
| 15 |
>>>> http://www.iozone.org |
| 16 |
>>>> |
| 17 |
>>>> It's in portage: "app-benchmarks/iozone". |
| 18 |
>>>> |
| 19 |
>>> |
| 20 |
>>> I installed it but trying to figure out how to use it. Jeez, what a man |
| 21 |
>>> page. O_O |
| 22 |
>>> |
| 23 |
>>> |
| 24 |
>> |
| 25 |
>> Yeah, it's a tough one. Also, very slow, assuming I understand it |
| 26 |
>> correctly. You have to use file sizes larger than the memory of the |
| 27 |
>> system, so on a 24GB system it takes (literally) a day or two to run. |
| 28 |
>> (Assuming I actually understood the man page!) ;-) |
| 29 |
>> |
| 30 |
>> - Mark |
| 31 |
>> |
| 32 |
>> |
| 33 |
>> |
| 34 |
> |
| 35 |
> I found this command and it worked pretty well. It does take a good while |
| 36 |
> to run tho. It wasn't to bad on my new rig but the old rig did take a |
| 37 |
> little while. |
| 38 |
> |
| 39 |
> iozone -R -l 5 -u 5 -r 4k -s 100m -F /home/f1 /home/f2 /home/f3 /home/f4 |
| 40 |
> /home/f5 | tee -a /tmp/iozone_results.txt & |
| 41 |
> |
| 42 |
> The results were much different than what hdparm shows. It shows the |
| 43 |
> 3Gbs/sec like they advertise they can do. I wonder which is more accurate? |
| 44 |
> :/ |
| 45 |
> |
| 46 |
> Dale |
| 47 |
|
| 48 |
Hi Dale, |
| 49 |
Good command, but it's missing one parameter that will likely make |
| 50 |
your results more meaningful. Try it again with the -e parameter |
| 51 |
added. I'll limit the tests also to just the initial write and read |
| 52 |
tests to give you something else to thiink about: |
| 53 |
|
| 54 |
iozone -R -l 5 -u 5 -r 4k -s 100m -i 0 -i 1 -e -F /home/f1 /home/f2 |
| 55 |
/home/f3 /home/f4 /home/f5 | tee -a /tmp/iozone_results.txt & |
| 56 |
|
| 57 |
I'll be more than happy to explain later why I suggested the |
| 58 |
changes, but in the spirit of someone's earlier comments about people |
| 59 |
doing research, give some thought to the results you get and see if |
| 60 |
you can explain to yourself why -e makes a difference. Do you believe |
| 61 |
the results you get are meaningful? |
| 62 |
|
| 63 |
If you want another good experiment, remove -e and make the files |
| 64 |
much larger, like 2g if you have an 8GB or DRAM. Again, explain to |
| 65 |
yourself why this makes a difference. If you do this experiment you |
| 66 |
_really_ should just do the -i 0 test only your first time through... |
| 67 |
;-) (HINT: You might want to open another terminal, run top and when |
| 68 |
in top hit iz and maybe m if you're not showing memory usage...) |
| 69 |
|
| 70 |
Cheers, |
| 71 |
Mark |
Archives