> -----Original Message-----
> From: octave klaba [mailto:[EMAIL PROTECTED]]
> Sent: Sunday, March 12, 2000 7:11 AM
> To: [EMAIL PROTECTED]
> Subject: Benchmarking
>
>
> Hi,
> I have begun some tests on my servers, but I have some stupid
> questions:
> - how can I get 200Mo/sec from SCSI-2 (please see my table) ?
> - if I test all servers with 100Mb file will it be correct ?
> or can I put the bigger size for the bigger system ?
> - the benchmark can be different for the different kernel ?
> 2.2.12 ou 2.2.14 ?
When doing benchmarks, you should test with a filesize AT LEAST 3 times how
much RAM you have. If you test with size=RAM, or smaller, than caching to
RAM will have a HUGE effect. Once you get to size=3xRAM, the effects of
caching are sufficiently reduced that it's no longer a sifignant factor. I
don't recognize that output, but you should also take a look at tiotest,
which is available from http://www.icon.fi/~mak/tiotest/.
Greg
