> > >territory.
> > >
> >
> > at high load generated by hundreds or more processes, especially when it
> > will be joint CPU and disk (and net) load, NetBSD should win without any
> > problems (unless it is not well configured).
> >
> > linux could possibly win in TCP transfer speed with one process.
>
> Once again, Wojtek, you're ready -- no, actually, downright *eager* -- to
> offer advice on a subject you don't actually know anything about.

if i do tests i don't know anything?

> SMP in NetBSD is currently "giant lock"; only one process can execute in the

i was talking about single CPU tests i did.

> kernel (e.g. perform I/O) at a time.  Though SMP in Linux sucks in many ways,

is one CPU for controlling I/O not enough?
it may be not enough maybe on 32-CPU machine but not here.


> it does not *always* suck in *this* way.  So, when you recommend using
> NetBSD in preference to Linux for I/O intensive configurations on a
> multiprocessor, you're giving pretty bogus advice.

anyway one CPU for driving I/O is enough.