Assuming you're running with diskd/aufs rather than ufs then I think
your numbers are alright. The default caching rules are quite permissive
and result in less caching than what might be possible, but more
"correct" caching.

You've paid for a support contract via Redhat - I suggest talking to
them about it. :)



Adrian

On Mon, Aug 06, 2007, Jan-Frode Myklebust wrote:
> I'm running squid (squid-2.6.STABLE6-4.el5) on an old IBM x330
> server (2x 1266MHZ PIII, 1GB RAM, 2 mirrored 36GB disks for OS
> and 20GB squid-spool), serving set-top-boxes' access to the 
> internet.
> 
> We have a feel for the proxy maybe being slow, but can't really
> pinpoint what the problem might be. Could somebody please have 
> a look at the below numbers to see if something stands out as 
> a potential performance problem ?
> 
> Some stats from calamaris for the last 24 hours:
> 
> Proxy statistics      
> -----------------
> Total amount:         requests        2818623
> unique hosts/users:   hosts           3427
> Total Bandwidth:      Byte            15607M
> Max. Bandwidth usage: MBit/sec        4.42
> Proxy efficiency (HIT [kB/sec] / DIRECT [kB/sec]): factor 10.45
> Average speed increase:       %               100.25
> TCP response time of 100%% requests:  msec    130
> 
> Cache statistics
> ----------------
> Total amount cached:  requests         2088373
> Request hit rate:     %               74.09
> Bandwidth savings:    Byte            8640M
> Bandwidth savings in Percent (Byte hit rate):         %       55.36
> Average cached object size:   Byte    4338
> Average direct object size:   Byte    10004
> Average object size:          Byte    5806
> 
> Incoming request peak
> ---------------------
> 429/second
> 6439/minute
> 238080/hour
> 
> 
> TCP-Request duration distribution
> -----------------------------------------
> msec          hits    byte
> <= 10         48%     23%
> <= 100        85%     60%
> <= 1000               99%     86%
> 
> File descriptors in use by squid are normally between 50-150, with peaks of
> 200-1000 once a week.
> 
> CPU-usage is seldom above 10%, with 1-2% iowait and less than 1% system cpu.
> 
> 
> 
>    -jf

-- 
- Xenion - http://www.xenion.com.au/ - VPS Hosting - Commercial Squid Support -
- $25/pm entry-level bandwidth-capped VPSes available in WA -

Reply via email to