On Tue, Mar 20, 2001, Florin Andrei wrote:
>
> Is it reasonable to grow cache_mem aggresively?
> I mean, suppose i have 2.5 GB RAM, and only about 500MB used by other
> processes. So i modified cache_mem to 200 MB. Now the main squid process
> is around 1 GB.
>
> The cache gets 60...100 req/second during the peak hours.
cache_mem is primarily used for in-transit objects, and for
"hot object caching". hot object caching isn't as neat as you
might think in squid-2 - you can twiddle with the usefulness
by changing maximum_object_size_in_memory. this defines the
threshold where squid will stop trying to keep the object in
memory and rely on swapping it in. If you've got oodles of
RAM you could try upping this to say 64k or 96k and see
if your memory hitrate improves.
>
> --
> Florin Andrei
>
Received on Wed Mar 21 2001 - 02:05:54 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:58:45 MST