[squid-users] Re: refresh patterns!

On Tue, 14 Aug 2007 20:30:28 +0200
Henrik Nordstrom <henrik@henriknordstrom.net> wrote:

> On tis, 2007-08-14 at 18:07 +0100, RW wrote:
>
> > Are refresh patterns very relevant to hit-rates?
>
> Yes, refresh_pattern tune how long objects is considered fresh, and
> also the tool to override HTTP freshness when needed..

Why does that affect the hit rate? If a browser makes a GET request
to squid for a stale object, and squid makes a GET-IMS request to the
server, and gives the client the object out of cache, then surely that's
a cache hit.

> > When an object becomes stale, squid will verify it on the next
> > access, which mostly results in a TCP_REFRESH_HIT, which is still a
> > hit. You
> > might argue that since it involves a round-trip it's a
> > "second-class" hit as far as latency is concerned, but I can't see
> > how refresh patterns have any significant effect on byte hit-rate.
> > Modern browsers also seem to be much more restrained in how they
> > reload, so I'm also a bit sceptical about reload-into-ims.
> >
> > I would have thought that most of the scope for improvement comes
> > from the cache acls.
>
> the cache acls can only further restrict what is cached, not make
> uncacheable content cacheable.

I'm talking about overriding the default for query urls. There's got to
be a better way than listing individual sites like youtube google-video
etc.

I was wondering how safe, and how useful something like this would be:

acl cache_upr urlpath_regex -i \.(png|jpe?g|gif|tif+|ico|css|js|swf|swv)($|&)
cache allow cache_upr
acl QUERY urlpath_regex cgi-bin \?
cache deny QUERY
Received on Tue Aug 14 2007 - 14:58:03 MDT