Well, first things first:
aufs means that squid won't delay processing during disk writes, other
than for cache hits on other requests.
If you are getting less than your full i/n connection - and you have no
other traffic, plus the sites you are accessing have enough grunt to
saturate your link, then you can review squid.
If you are getting good performance (say 80% of your links nominal speed
during busy periods), then there is unlikely to be a bottleneck in your
squid environment.
From here, if there is a bottleneck (the first case above), then
identify the cause of that. If there isn't, then you can look at
optimising for cost (greatest byte count hit %), performance (greatest
request hit % and lowest median service time)...
Generally though, you shouldn't need to twist too many knobs.
As for not caching large / streaming files other than windows update -
thats something that GDSF will address anyway. Unless you identify a
disk performance bottleneck as part of troubleshooting a performance
problem, I would keep the environment as simple as possible. That gives
you the greatest flexability.
I.e. when XYZ vendor sends out an email with a link to a mp3 for a new
ad campaign, all your marketing folk *will* go to the same .mp3 URL :}.
Rob
-- GPG key available at: <http://users.bigpond.net.au/robertc/keys.txt>.
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:15:36 MST