>> 1)
>> A = disk partition
>>
>> time dd if=A bs=32k of=/dev/null count=5000
>>
>> linux box will use 90% of the cpu, freebsd will use 2%
>> [given a decent scsi system]
GR> Are you crazy ?
GR> Here is what I get on a PII-233 under Linux and a decent SCSI system:
I agree that the time dd ... test is not very good..
GR> And I have given a try under FreeBSD on the same system but didn't had
GR> time to move the results to Linux world:
GR> Using block device: 6.4 s CPU reported and 8.5 MB/s
GR> calculated. Using raw device : about no cpu reported (0.3s)
GR> and 18 MB/s calculated.
GR> I may report you the exact values if you need to be convinced,
GR> but if you have a decent SCSI system, you may try it by
GR> yourself.
GR> Results on my system
GR> - ---------------------
GR> - - Using block device FreeBSD is way slower than Linux.
GR> - - Using raw device it is as fast and about no CPU is
GR> consumed as expected
GR> by not stupid people.
I do not agree. I've run lots of benchmarks on both Linux (RH5.1 both
with 2.0.35 and 2.1.xxx kernels) and on FreeBSD-current, using
fast-SCSI-II and UDMA disks.
I used benchmarks such as bonnie that mearure both sequential
throughput, and random access, and also the CPU usage caused by the
tests.
FreeBSD was faster in all cases, sometimes a bit, sometimes a lot.
For example, my new UDMA disk got 12MB/s with 2% CPU on FreeBSD, on
Linux I got 7MB/s with more (about 15%) CPU usage.
Especially, while doing such test with a lot of disk I/O, FreeBSD
remains just as snappy for interactive response as if the system were
idle, on Linux doing heavy disk I/O, the interactive response clearly
deteriorates.
Still, I'm going to install Linux again (RH5.2) because of
applications (oracle-8, soon JDK with JIT). It is really a pity that
the most superior and best organized system is behind
w.r.t. commercial support.
-- /\_/\ ( o.o ) Peter Mutsaers | Abcoude (Utrecht), | Trust me, I know ) ^ ( plm@xs4all.nl | the Netherlands | what I'm doing.- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/