HI Amos
On 30/10/2013 2:51 a.m., Ayham Abou Afach wrote:
> Hi
>
> i have the folloing problem after moving from squid 3.1 to ( 3.2 or 3.3
> ) with same config
> bandwidth saving decreases to about 50%
> what is the deffirance between versions related to caching behaviour ???
>
> any one has a solution to this problem...
>
> Regards
> On 10/29/2013 11:38 PM, Amos Jeffries wrote:
> The big caching related changes:
>
> * 3.2 version is now HTTP/1.1 - with extended cacheability and
revalidation behaviour.
> - In some cases HTTP/1.0-based savings calculation can show a
decrease even as total bandwidth is reduced.
> - More cacheable content (in HTTP/1.1 almost anything is cacheable)
can mean more spare HIT rate.
> - avg store object age, size, near-HIT ratio also need to be
considered more important for HTTP/1.1
> - NP: Several of the refresh_pattern ignore-* and override-* options
cause *reduction* in HTTP/1.1 compliant caches.
So that means that the traffic will be mor validated in HTTP/1.1 thats
why we are losing some objects to be cached
i am using refresh patterns without any effect
> * 3.2 version is validating intercepted traffic safety.
> - Unsafe traffic will not be cached.
all my traffic is passing to my 3.2 box as spoofed traffic with tproxy,
but what do u mean by unsave traffic ???
> * 3.2 cache size calculations have been updated.
> - Uncovering a bug where the maximum_object_size directive must be
placed above the cache_dir for it to have any effect raising the object
size limit on those cache_dir.
> Amos
i am aware for this bug and have 2GB objects in my cache and i think it
is the limit by squid even u set it for 4 G in config
what we should know is there any config should be added to 3.2 config
file to compensate the reduction in bandwidth saving ????
i have 4 caches 3 of squid 3.1 and the 4th one is using squid 3.2; i
tried 1 worker and 4 workers and 3.1 versions are caching almost twice
Thanks
Ayham
