[squid-users] Simultaneous requests in accelerator mode

Hello everybody,

I had to accelerate a high traffic webserver with lots of very very
enormous PHP-scripts, I took squid in accelerator mode with suitable
refresh_patterns, and it did a really good job - the load on the webserver
decreased significantly.

However - its job could be much better. :-)
When requests arrive simultaneously and the corresponding cache entries
are outdated, squid forwards these requests simultaneously to the
webserver - and this behavior is still really bad for the webserver in
peek times ...

With my understanding, it would be much better, if squid would recognize
it's already reloading the request from the webserver and answer all other
requests with the "old" cache entries until it has the new ones.

I couldn't find any information on this problem, only one post to this
list (http://www.squid-cache.org/mail-archive/squid-users/200110/1089.html)
which unfortunatly stood unanswered ...

I fear my wishes do conflict with the philosophy squid was written and
there is no easy way to implement it - but before I have to start hacking
I better ask first this list ... :-)

rolf