On Wednesday, August 08, 2001 07:07:39 PM +0400 "Vladimir V.Saveliev"
<[EMAIL PROTECTED]> wrote:
> Hi
>
> "Vladimir V.Saveliev" wrote:
>
>> >
>> > dbench 32
>> > Throughput 34.7552 MB/sec (NB=43.444 MB/sec 347.552 MBit/sec)
>> > 24.620u 73.980s 2:02.55 80.4% 0+0k 0+0io 911pf+0w
>> >
>> > Throughput 21.8827 MB/sec (NB=27.3533 MB/sec 218.827 MBit/sec)
>> > 25.410u 76.610s 3:14.04 52.5% 0+0k 0+0io 912pf+0w
>> >
>>
>> Hmm, dbench 32 does not show this big difference for me.
>> 2.4.7-ac3: Throughput 9.1809 MB/sec (NB=11.4761 MB/sec 91.809 MBit/sec)
>> 2.4.7-ac4: Throughput 8.60357 MB/sec (NB=10.7545 MB/sec 86.0357
>> MBit/sec)
>>
>
> Sorry, forgot to mention - this is dbench 1.2
>
> Just got results of 'dbench 32' for ext2:
> 2.4.7-ac3: Throughput 10.2969 MB/sec (NB=12.8711 MB/sec 102.969 MBit/sec)
> 2.4.7-ac4: Throughput 9.24066 MB/sec (NB=11.5508 MB/sec 92.4066 MBit/sec)
>
> It looks like ac4 touched performance of ext2 a bit as well.
It is pretty hard to get good consistent dbench speeds, especially when the
run takes less than 10 minutes or so. The problem is that
bdflush/kupdate/kswapd can trigger at different times during the run (or in
some cases not at all), giving different numbers from run to run.
The neat thing about dbench is that it runs faster if the FS is unfair.
When all the processes finish at the same time, you get the worst numbers.
-chris
