Re: os io unit , db block size relation

utkanbir wrote:
> Hi Fabrizio ,Yong
>
> Sorry for delay , it will take some time to complete
> the other jobs. Based on your comments and tests , i
> made some tests against ocfs so i believe this time
> each time i read data from disk no caching is
> occurred.
>
> This is iostat output:
>
> avg-cpu: %user %nice %sys %idle
> 260.95 0.00 3.90 759.15
>
> Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s
> avgrq-sz avgqu-sz await svctm %util
> sde 6360.80 37.20 145.60 3.20 6542.40 40.40
> 44.24 1.25 8.41 2.31 35.26
>
> While your tests show great difference between rrqm/s
> and rsec/s or wrqm/s and wsec/s , here these values
> are very close to each other.
>

You are doing 512 byte reads and writes.

Quoting from ocfs2 documentation:

"OCFS only supported blocks of 512 bytes. OCFS2 supports a blocksize of 512B, 1K, 2K, or 4K. The maximum is 4K because the minimum cluster size is 4K."

So I believe you are using OCFS(1).

Unfortunately I don't have any ocfs installation. I'll try to set a test environment tomorrow.

About iostat: have a look at
http://fxr.watson.org/fxr/source/Documentation/iostats.txt?v=linux-2.6.0-test9

The value of iostat are simply taken from the /proc filesystem and averaged per second. You can query the source (with cat) instead of a tool...

-- 

Fabrizio Magni

fabrizio.magni_at_mycontinent.com

replace mycontinent with europe