Subject: Re: Simple thought...
To: Frank Kardel <Frank.Kardel@Acrys.COM>
From: Greywolf <firstname.lastname@example.org>
Date: 06/10/2002 10:15:57
On Mon, 10 Jun 2002, Frank Kardel wrote:
# With UBC there is also another interesting parameter generating much
# joy when increased:
# sysctl -w kern.maxvnodes=64000 (for bigger machines arount 700MB RAM)
# The data-pages now hang off the vnodes in UBC if i read the hints everywhere
# right. Now the more vnodes you have the more data can be tacked onto them.
# Increasing the buffer cache only helps meta data and takes memory away for
# vnode data caching. Try to look at the "vmstat buf" statistics page and look
# for vnode data pages. My machine (768MB) usually caches around 500-600MB
# file system data. And more - lingering vnodes do not need to be filled
# from the buffer cache. Effects are that kernel ld-runs are completely filled
# from memory (just written .o files) and the bufcache is only used for
# moving meta data. Set me right if i overlooked something here.
I thought vnodes WERE metadata. Or is it that the _i_node data is the
stuff (under the vnode) that gets cached and the vnode is just a pointer,
and there's no cache for that?
[I seem to recall that vnodes did get cached; they were the one parameter
that if they maxed out (according to pstat) you didn't have to really
worry about it (unlike files or swap space)...]
Is there a tool that displays buffer cache statistics in some form?
I don't think pstat is what I want for that (although it does show the
open file table...), and I don't think vmstat is what I want, either...
[It would be interesting to see the buffer cache's references. Useful?
Probably not on a large scale. But as I'm not well-versed in kernel
hacking, I'm curious to see what's there...]
# Frank Kardel
NetBSD: Microsoft ask you where you want to go. BSD gets you there.