Re: Direct as last resort

From: Reuben Farrelly <reuben.farrelly@dont-contact.us>
Date: Sat, 31 Jul 1999 16:13:41 +1000

Hi,

At Friday 30/07/99 07:38 PM -1000, Clifton Royston wrote:
>Antony writes:
> > At the risk of annoying the regulars, can anyone enlighten me on how to
> do the
> > following...
>
>Dumb question here: why wouldn't you *want* this to be the case? If
>the parents end up responding faster to ICP, they're probably going to
>deliver the hits faster too. Isn't that what you want?

I am guessing that he wanted to use the siblings in preference to the
parents because traffic obtained by the siblings is cheaper (?)

>It may be technically sweet to have a fully distributed network with
>load going off to the siblings, but if the parents serve the hits
>better, it would seem to me you ought to just let Squid use them. The
>sibs would still be useful in that they'd end up picking up hits if the
>parents were saturated or down for some reason.

Which brings up an interesting point. I thought that both parents and
siblings could be weighted via the cache_peer directive. Closer reading of
the squid.conf file shows that weightings are only relevant to parents and
not siblings.

Is there any reason or situation why squid only allows weighting of parents
and not siblings as well? (an idea for a patch, someone? ;) )

Reuben
Received on Sat Jul 31 1999 - 05:07:04 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:47:38 MST