cachemgr - improvement?

Hi,
  excuse me if this has been raised before, but I don't remember seeing
it mentioned anywhere.

Given that everyone I know or have talked to about squid is more
concerned about bandwidth and byte counts, is it possible or practical
that changed be made to reflect "Hit Rate" as a percentage of BYTES
rather than percentage of HITS?

For instance, I've now been running stats on bytes-to and bytes-from my
proxy, and found it's more than 12% in error *on volume* (which is all
I'm interested in). Really, who gives a damn if 50% of all FILES
requested were in the cache, if the other 50% that missed were all
megabyte files?

I know, there are scripts there to do this analysis, but for my money,
cachemgr is a nice, tidy, quick, convenient way to keep an eye on several
proxy machines without all the mucking about. The accountants only want
to see the justification for a server, and it's easy to say from the
stats "55% hit on 50Gb of transfer = $xxx". Presto, justification.

If the 55% happened to be all 3Kb thumbnail images, and the other 45% of
misses were the 800Kb photos, things are decidedly worse!

So, without knowing what's involved, what's the opinion on this? Is there
some fundamental reasoning I'm missing? I suppose, statistically, over a
large enough sample, the two percentages will converge, but I'd
personally like to see it the other way, and those I've spoken with
about it agree.

Regards, RossW
Received on Thu May 08 1997 - 05:35:54 MDT