On Sun, 10 Feb 2002, David Laight wrote:

# Simas Mockevicius wrote:
# >
# > Hmm,
# >
# > here is a lot of questions, to find out what we need to test ?
# > :) Network, pheripherals, devices and etc...
#
# System call entry exit for 'nop' calls?
#
# My 'gut feel' is that the filesystem performance doesn't match other
# systems I've used (Unixware 2/7, Solaris etc).  I think there are
# excessive disk activities, probably caused by file system metadata not
# being cached very well.

I think this is probably a side-effect of the UBC code.  I remember
the first time I did a tar copy of an entire SunOS CD, and aborted half-
way through it, I went back and re-did it (to extract to the proper
directory this time), and it seemed to have cached a lot of the
metadata.  This was NetBSD/sparc 1.1.

As far as I can tell, there's a few performance issues to be worked
out with regard to the metadata caching, and unless I'm totally
mistaken, it is being worked on.

# 	David


				--*greywolf;
--
NetBSD: Rock solid!