Hi Grendel,
> The hdparm command itself includes switches to benchmark your hard disk,
> try "hdparm -T /dev/hda" and "hdparm -t /dev/hda"
I know this. The -T option is what I am not looking for, and -t is not
usefull for all cases. For example, issueing the dd command I can see a clear
improvement of speed when setting the filesystem read ahead (-a) to a high
number, lets say 32, but this improvement is not visible in the hdparm test.
But I want to test if increasing the filesystem read ahead buffers slows down
ls, because that would be unwanted. So I want to compare settings for
different ways of accessing the drive. That is why I want to be able to do
some custom tests. And I must say that the results from the dd command are
pretty stable (not to much variation).
> when you run ls for the second time it
> will instantly deliver the cached result from memory, thus giving a much
> faster result the second time.
This is just the thing I want to avoid.
> It must be possible to disable read and
> write disk caching for linux, but I dont know much on how to accomplish
> this.
I do not necessarily have to disable caching, but I want to be able to flush
the directory cache. The disk cache can be flushed using hdparm -f, and as I
said, this makes the results from dd consistent. But it seems the kernel
caches the directories in a different way, so issueing ls a second time speeds
up the result, even after issueing a hdparm -f.
> Would hdparm benchmarking not meet your requirements? Its easier and the
> reliable IMHO.
As explained above, no.
Thanks anyway,
Leonard.
_______________________________________________
Redhat-list mailing list
[EMAIL PROTECTED]
https://listman.redhat.com/mailman/listinfo/redhat-list
