Re: TCP_HIT in Squid 1.1.X is the same as in Squid2.0??

On Wed, 7 Oct 1998, Javier Puche. CSIC RedIRIS wrote:

> Hello all,
>
> After migrating from Squid 1.1.21 on a Sun Ultra 2 Solaris 2.6 to
> Squid 2.0 on the same platform and experiencing an amazing speed up of
> the cache, I do not understand why the output of calamaris.pl (v 1.20)
> shows TCP_HIT latencies quite higher than before, even though the user
> perception does not correspond to that.

TCP_HIT in Squid 2.0 is the same as in Squid 1.1, I guess. There are new
types of hits though if you are using Cache Digests. I do not know if
calamaris can automatically detect new tags in the access log.

Apart from that, how does Calamaris measure the "average" transfer speed?
Does it use means or medians? If means are used, the delay- or time-related
stats tell you pretty much nothing (regardless of Squid version) because of a
huge heavy tails of delay distributions.

The correlation between median delays and user perception of "speed" is the
third, still unanswered, question!

$0.02,

Alex.
Received on Wed Oct 07 1998 - 10:57:52 MDT