Re: HIT rate first level Web cache compared to Top-level Web cac

From: Jan Torreele <>
Date: Tue, 2 Sep 1997 17:14:47 +0200 (MET DST)

Hi all,

I brought forward this point already earlier in the list. We should first
define what a HIT is, before starting to compare. The hit rates reported by
cachemgr, squidstats, squidtimes, spa and other log file analysis scripts
vary wildly (from 10% to 40% for the same day). Shouldn't we all agree on
using one simple awk or perl script to calculate the hit rate so as to
assure that we all talk about the same thing? Alternatively we could
all agree upon using the hitrate as reported by the cachemgr (btw. can
somebody tell me how cachemgr calculates its hitrate?)


Jan Torreele				E-mail:
BELNET Service Support Team		Tel.:	 +32 (0) 2 / 238 34 70
On Tue, 2 Sep 1997, Michael Pelletier wrote:
> On Tue, 2 Sep 1997, Adam Neat wrote:
> > >  - first, note that 50% is very close to the optimal cache hit rate.
> > 
> > How are you able to get such high hit rates? We have 3 proxy servers running 
> > wiht Squid 1.1.11 and we only get between 12% and 18% hit rates!
> > 
> > What are we doing wrong?
> Perhaps we should come up with a different terminology for object hit
> percentages and bytes-saved percentages.  I suspect that 50% refers to the
> number of objects found in the cache regardless of their size, and 12-18%
> refers to the percentage of bytes transmitted from the cache.
> 	-Mike Pelletier.
Received on Tue Sep 02 1997 - 08:24:07 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:36:54 MST