Ted Lemon <mellon@isc.org> writes:
> > Wouldn't result this in a serious performance loss?
> 
> I'd love to see profiling numbers, but I can't see any reason why it
> would.  In general, the cost of swapping is going to be lost in the
> pipeline because of memory latency - only on architectures with no
> pipelining or _really_ fast memory is this likely to make a
> difference.

Even ignoring memory latency -- consider the eternity (compared to
memory latency) it takes to write a block out to disk...

Perry