[TriLUG] Squid proxy guidelines

[Jon Carnes]

Dan, those are excellent stats.

I run a squid cluster for a large government client and we average about
a 25% hit from cache - which my clients are extremely happy about.

>From what I've seen in industry, anything above 20% is good.

Jeremy - you generate these stats from the log files. There is a perl
program that comes with Squid that lets you parse the log file for
exactly the stats shown below...

I've written my own parsers in Python.  They let my clients run various
web usage reports (for individuals and for groups).

Jon

On Wed, 2004-09-01 at 12:33, Dan Monjar wrote:
> Who here is using Squid?  I'd like to compare stats to figure out if I need 
> to tune this thing some more.  For instance:
> 
> -------
> Connection information for squid:
> Number of clients accessing cache:454
> Number of HTTP requests received:217911
> Number of ICP messages received:0
> Number of ICP messages sent:0
> Number of queued ICP replies:0
> Request failure ratio: 0.00
> Average HTTP requests per minute since start:148.8
> Average ICP messages per minute since start:0.0
> Select loop called: 9105362 times, 9.653 ms avg
> 
> Cache information for squid:
> Request Hit Ratios:5min: 50.5%, 60min: 40.6%
> Byte Hit Ratios:5min: 21.0%, 60min: 17.3%
> Request Memory Hit Ratios:5min: 10.2%, 60min: 24.7%
> Request Disk Hit Ratios:5min: 6.4%, 60min: 14.2%
> Storage Swap size:1887434 KB
> Storage Mem size:115308 KB
> Mean Object Size:9.75 KB
> Requests given to unlinkd:34552
> ----------
> 
> Is a 5 minute hit ratio of 50% good or bad? Of course I'd like 100% been 
> then I'd have to cache the Internet.  what type of utilization and client 
> counts are you seeing?  How big is your RAM cache and disk cache?  I'm 
> interested in discussing stuff like that...
> 
> -- 
> Dan Monjar