We have just changed from CERN to Squid. CERN is buggy and slow and Squid
is much nicer so we would very much like NOT going back to CERN. But...
1. [I know nothing about HTTP-protocol. I got the info below from my boss.]
Squid caches dynamic pages! This is because very few pages are marked
correctly. If the server answers "HTTP/1.0 201 Created" instead of
"HTTP/1.0 201 OK", the page is dynamic and should not be cached. But
we haven't found one single server that answers like this. Otherwise
this would work, we put this "Created"-header in a few of our dynamic
pages and then Squid doesn't cache them. Since we can't rely on servers
to handle this, dynamic pages must be correctly marked. But very few
are. CERN handles this by checking whether "Last-modified"-header exists
or not. If it exists, the page is static, otherwise dynamic. This has
worked fine.
So, am I totally wrong here? In that case, please help me or we will
have to consider going back to CERN and I don't want that.
If I'm not wrong and this is normal Squid behaviour, what can be done?
2. Is Squid a cpuhog or have we configured it wrong or maybe to small machine?
Some of our customers have not yet had time to alter their settings
and are still going through CERN. Squid is presently doing 25% of the
proxy/cache-work, CERN does the rest. But Squid uses twice the cputime
compared to CERN.
The Squid-process is 8 MB + 20 MB swap = 28 MB. (CERN-process is
500 kB...) The machine has 32 MB memory, and 32 MB swap is used.
We know that Squid uses a lot of memory, we will upgrade. Could this
be the reason that it uses a lot of cputime?
Thanks for any comments on the above!
Peter Olsson pol@leissner.se
Received on Mon Aug 19 1996 - 02:34:53 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:32:49 MST