On Fri, 23 Feb 2001 wojtek@wojtek.from.pl wrote:

> >
> > 	Chuck could answer better than I, but I imagine that the gain from
> > 	making metadata buffers dynamic is not significant enough to make
> > 	it worth working on at this point. Maybe when all filesystems
> > 	perform close to optimally under UBC and the automatic tuning
> > 	works fully then it might be time to address the static metadata
> > 	cache.
> OK but why it's still same amount of memory allocated for buffers? it
> should be smaller.
>
	I'm of the firm belief that the amount of buffer cache previously
	allocated was too small in most cases, so leaving that same amount
	for metadata is a good thing, but I agree it could safely be
	reduced somewhat.

		David/absolute		-- www.netbsd.org: No hype required --