analyze hitrates by type of object

I'm running a 3Gb squidcache on DEC/Alpha, direct parented behind a much
larger resource. 100 users at most. Nothing heavy.

I'm reasonably content that I only achieve around 14% hitrate, the re-use
of the parent cache achieves my traffic cost-savings by re-directing to
domestic tarrifs, so I'm really ahead even when I miss. But should I be
so complacent?

But looking at the graphs, I am struck by how *variable* the hitrate is
by type of object. Great for images, terrible for <other>, so-so for .html

I am also struck by the incidence of .class and other objects I infer from
little evidence are now 'out there'

Can anybody more clueful comment on:

        a) the achievability of high hitrates en masse for the small-site
                 cache: where do we say 'its working'

        b) the variance of hitrate by type of object: why might things vary
                so much depending on the MIME code?

        c) should I care? In particular, should .class instances and .jar
                 instances get more attention somehow?

-George

--
George Michaelson         |  DSTC Pty Ltd
Email: ggm@dstc.edu.au    |  University of Qld 4072
Phone: +61 7 3365 4310    |  Australia
  Fax: +61 7 3365 4311    |  http://www.dstc.edu.au