Jochen Keil writes:
> Just an update to my last email
> 
> Jochen Keil wrote:
> > In my opinion the problem has something to do with RAIDFrame. I'll try 
> > to get Vinum to run with NetBSD because my last half-hearted try wasn't 
> > too successful.
> Vinum is initialising right now and it seems that the tcp/ip stack of 
> netbsd doesn't like me..
> No luck so far, ping response times are partially even worser than with 
> RAIDFrame. This is very disappointing for me.
> Maybe i should send a PR?

As Manuel suggests, try pushing the disks without RAIDframe or vinum.
Maybe even with something like:
  
  foreach i (0 1 2 3)
   dd if=/dev/rwd0d of=/dev/null bs=1m &
  end

and then try the 'ping test' again.

It's my guess that the Promise Ultra100TX2/ATA is killing your 
machines with interrupts... :(   (running 'systat vmstat' while 
pushing the disks might give you an indication of what's happening in 
interrupt land.)

I've certainly not seen the problem you're describing even when 
pushing RAIDframe under heavy IO load.

Later...

Greg Oster