Matthias Scheler wrote:
> round-trip min/avg/max/stddev = 0.174/0.195/0.211/0.016 ms
> 
> That's two (loaded) NetBSD boxes linked over wm(4) interfaces via
> a gigabit ethernet switch.
There's no switch in my setup. The other box is running freebsd and has 
a cheap noname 100mbit nic. The reason why i'm looking at netbsd is that 
the problem mainly occurs when using raidframe.

>>-Using a software raid with raidframe makes it even worse
> 
> 
> The sender used above is using RAIDframe.
> 
> Can you repeat the tests without using cgd(4), please?
Nearly all my test were performed without cgd because i realised pretty 
fast that cgd is yet another source of distortion and it even regularly 
crashed the box. With this mail i compiled a kernel without cgd but it 
didn't make any difference.

Out of curiosity i tried a RAID0 setup with two disks but theres no 
difference to the previous RAID5 setup concerning my problem so far.
Is there a way to optimise/prioritise raidframe or the tcp/ip 
implemention in the netbsd kernel? I'm not very well versed with this 
stuff..

Best Regards.