It depends on the number of disks thats you use for the cache on disk.
as a rule of thumb: 10 I/Os per disk is fine, so 10 threads per disk.
Only if you use very high performance disk arrays you may
increase the number of threads per (logical) disk.
Marcus
J. Webster wrote:
> Would this dramatically improve performance or it it best left at default?
>
> ----------------------------------------
>> Date: Tue, 9 Feb 2010 17:01:46 +1300
>> From: squid3_at_treenet.co.nz
>> To: squid-users_at_squid-cache.org
>> Subject: Re: [squid-users] DNUMTHREADS
>>
>> J. Webster wrote:
>>> Is it recommended to recompile squid and increase the DNUMTHREADS value?
>>> I read that 30 could easily be used on a 500MHz machine and my machine is more than 2GHz so would it give an improvement to squid performance.
>>> I have been reading through this document here, which recommends various changes including using the reiserfs filesystem.
>>> My machine is CentOS.
>>>
>>> http://blog.last.fm/2007/08/30/squid-optimization-guide
>>>
>> Not sure how he got that info Squid provides the ./configure
>> --enable-async-io[=N_THREADS] option as far back as I can see.
>>
>> It only affects AUFS disk storage.
>>
>> Amos
>> --
>> Please be using
>> Current Stable Squid 2.7.STABLE7 or 3.0.STABLE23
>> Current Beta Squid 3.1.0.16
>
> _________________________________________________________________
> Do you have a story that started on Hotmail? Tell us now
> http://clk.atdmt.com/UKM/go/195013117/direct/01/
>
Received on Tue Feb 09 2010 - 22:13:56 MST
This archive was generated by hypermail 2.2.0 : Wed Feb 10 2010 - 12:00:05 MST