[squid-users] LFUDA replacement policy - Byte Hit 18%

From: Martin Marji Cermak <mc1@dont-contact.us>
Date: Thu, 09 Dec 2004 16:19:45 +0800

Hello Guys,
I'm trying to save some bandwidth, so I have

   cache_replacement_policy heap LFUDA
   maximum_object_size 200 MB

in my config file, because I believed this policy was the best one for
maximising the Byte Hit Ratio.

I had to wait for 3 days to get my cache full (84 GB total), because I
started with an empty cache (my environment: small ISP; 3500 ADSL
customers; average Client Http Request: 119 req/s, 205 req/s peek;
CacheHTTPout: 1MB/s, 2 MB/s peek).

During these 3 days, before the cache got full, my peek time Byte Hit
Ratio (60min) was over 20% (max 23%).
But, after the cache had filled, my peek time Byte Hit Ratio (60 min)
was between 16% and 18% only!

I guess, because the cache had filled, Squid started replacing instead
of adding pages. I guess Squid must have released some files which were
quite large or/and frequently used.

I'm surprised about that - I didn't expect the Byte Ratio going down
after the cache had filled.

Any comments of this?

Cheers, Marji

PS: I will post more info about this when I have more stats.
Received on Thu Dec 09 2004 - 01:19:13 MST

This archive was generated by hypermail pre-2.1.9 : Sat Jan 01 2005 - 12:00:02 MST