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.
