Re: [squid-users] cache refresh

From: Alex Rousskov <rousskov@dont-contact.us>
Date: Sat, 11 Aug 2001 11:18:46 -0600 (MDT)

On Sun, 12 Aug 2001, TomC wrote:

> But, some benefit too as I will no need to waste the bandwidth to
> refersh the copies in cache in peak hours

Only if those copies are still fresh. How many objects expire before
your script is run but stay fresh for the next 12-20 hours?

Also, if the object in question is popular, the default scheme will
yield just one extra miss while still producing many hits so your
overall hit ratio is unlikely to change much. I suggest that you
estimate or, better, measure the savings before you start believing
that the savings are worth the expenses.

Alex.
 
> Alex Rousskov wrote:
>
> > On Sat, 11 Aug 2001, TomC wrote:
> >
> > > Hi all,
> > > is it possbile to refresh the whole cache in non-peak hours by a script:
> > >
> > > grep all the url from the accesss_log
> > > then use a client -r -p 8080 -h host url
> > >
> > > Then, the cache will be updated automatically?
> >
> > Yes, however:
> >
> > - many origin servers will not like your aggressive fetches
> > and may block you
> > - you will request many uncachable objects
> > - you will reload objects that were cached and fresh
> > - many reloaded objects will still expire after/while your
> > script is running
> > - you will get many errors because your script does not use
> > custom headers such as authentication that real clients
> > were using
> > - this will create a lot of "extra" traffic on the network
> > - in most cases, Squid's default behavior will be nearly
> > optimal (or, at least, better than your script)
> >
> > Alex.
>
>
Received on Sat Aug 11 2001 - 11:18:53 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:01:34 MST