Re: hit rates and maximum_object_size

Neale Banks wrote:

> > > Storage LRU Expiration Age: 7.96 days
> > > Mean Object Size: 11.25 KB
> > > Requests given to unlinkd: 0
> >
> > If this has been running for some days, then you do not have need for
> > much more disk.
>
> OK, any suggestions as to the consensus for "reasonable" bounds on this
> and any related measures?

I would say a LRU age of about one week is quite well balanced for
medium sized installations. More is probably not worth the effort. This
is assuming that your cache has filled up and then handled about the
same amount of traffic it took to fill the cache, and that your Squid
has been running some time with nominal load when you take the measure.

This is a subjective opinion based on a number of things, and I have not
measured it myself in real life.

> That would be:
>
> Squid-2.2.DEVEL3: Range request could cause bandwidth spikes
> Range requests to servers/objects not supporting range requests could
> cause bandwidth spikes and/or negative hit ratio.
>
> Squid-2.3.DEVEL3: HEAD and ftp://...
> Support HEAD ftp://.. requests without fetching the whole object.

Right, except for a minor versioning error (web page will be corrected
on next update).

Note to other readers: The version number listed next to my patches is
irrelevant for most uses. What is relevant is under which heading on my
Squid page the patch is listed.

> Given that I am currently running 2.2.STABLE4, I should be upgrading to
> STABLE5 first?

Since you need to recompile Squid anyway you might just as well move to
2.2.STABLE5 at the same time to get all NLANR bugfixes rolled into one.
Note however that dstdomain based ACL's have some problems in STABLE5 so
if you use them then you may need to be careful.

--
Henrik Nordstrom
Squid hacker
http://hem.passagen.se/hno/squid/