Re: [squid-users] Most efficient cache method

From: Chris Robertson <[email protected]>
Date: Thu, 11 Jan 2007 15:36:27 -0900

John Halfpenny wrote:
> Hi all.
>
> I have 3 squids running on one server:
>
> A ----\
> * ---- C
> B ----/
>
> A and B are for NT authent, * denotes dansguardian, and C is the squid for dansguardian to run on.
>
> Would I be better off speedwise caching at A and B or at the parent C? Or at all three? :-)
>
> Thanks for the endless support
>
> John
>
>
>

It seems to me, there is not enough data to give a single valid answer.

Assuming the box is not being stressed by the Dan's Guardian filtering,
you would be best served by caching on "C". Cache able requests made
through "A" will be served from cache when the same request is made
through "B". Requests that are not allowed by either will not be
cached. Requests that are only allowed to "A" will still only be
accessible from "A". Cached content will only be stored in one place
("C's" data store) making the best use of space.

If DG is putting undue strain on your CPU, AND you have a slow/expensive
link to the outside world, cache on all three. Fewer requests will hit
DG, and those that do will have a greater chance of being cached on "C"
by the partner auth server's request. This setup will cache the least
amount of unique content.

If DG is putting strain on your CPU but you have a fast link to the
outside, only cache on "A" and "B". That will even further reduce the
load on DG (as the space allocated to "C" would be split among "A" and
"B", giving them better hit ratios).

Chris
Received on Thu Jan 11 2007 - 17:36:35 MST

This archive was generated by hypermail pre-2.1.9 : Thu Feb 01 2007 - 12:00:01 MST