[squid-users] Multiple downloads of the same URS in accel mode.

Hi,

I am stuck;-) I went through the documentation, I read the FAQ, I searched
the list archives but I can’t figure out any solution to my problem. Maybe I
am looking for the wrong stuff or I simply misunderstood the matter
completely.

We are running SQUID in accelerator mode for a webserver which produces
dynamic content. It catches about 80% if the incoming requests. We control
the caching of these Objects with ‘CacheTime=Xs;’ in the HTTP header. So the
cache will get a new copy of the object every X seconds.

If the system usage increases in peek times, the webserver is operating at
its performance limit to satisfy the caches need of fresh cacheable objects
and uncacheable objects. It seems to me that something strange happens:

A client asks for an expired object therefore the cache tries to get the
object from the webserver.

The webserver queues this request and needs 20 seconds to deliver the
output.

Meenwhile another client requests the same object. SQUID can’t find a valid
object in the cache and seems to request the same object a second time.

This happens again and again since not only 2 clients are requesting this
object during this 20 seconds.

So the webserver looses its ‘protection’ and is suddenly flooded with
requests. That has, of course, an extremely negative impact on the response
time (if not killing it completely) so the next object that expires will
produce the same effect.

Isn’t there a way to tell squid to pull the fresh object once and queue all
pending requests for it until it was successfully received? Or even better:
Send out the existing but stale object to the client until if finished
getting the fresh one?

Any hint would be greatly appreciated

Thanks

Thomas Ristic

---
Windows 9x: a 32 bit extensions and a graphical shell for a 16 bit patch to
an 8 bit operating system originally coded for a 4 bit microprocessor,
written by a 2 bit company that can't stand 1 bit of competition...