Delaying fetches

Hi all,

Our squid cache is fetching around 15-20% more data than access.log reports.

I detected this by comparing the total "external fetches" summarised by the
"calamaris.pl" tool against the actual MB's sent to squid by external sources
(using ipfw and netflow).

I assume that our clients are aborting requests after squid has fetched
the whole object from external sources. I understand that squid's
CLIENT_ABORT
logs the data sent to the client, not that fetched, hence the discreprency
between log summaries and actual fetched data.

Can anybody suggest a means of "slowing down" squid's fetching of external
objects? Ideally it would keep only 10-15 KB ahead of the client.
In Australia we pay for traffic received (ie 19c/MB).

If anybody is willing to write a patch for us, please e-mail me (we will
happily reimburse you if necessary). We are willing to make that patch
freely available.

Regards

Greg Holloway
Received on Fri Oct 10 1997 - 07:00:08 MDT