Regarding BIND9.3.2

[Marco Schumann]

Marco Schumann schrieb:
>> Hello,
>>
>> how does the system malloc influence the performance on multi processor
>> machines? Are there known side effects when I use "--disable-threads"?
>> Are these contrary options?
>>
> Hello,
> 
> to be more specific I need to reference my own post from Feb 15.
> (Resolver (bind-9.3.2) eats processor)
> I see a strange behaviour of bind-9.3.2 resolvers which has been
> verified with bind-9.3.1, too.
> 
> I compiled with system malloc and "--disable-threads", occasionally the
> named process uses 100% of CPU over hours answering the queries very slowly.
> 
> Currently I am testing with "ISC_MEM_USE_INTERNAL_MALLOC 1", so I still
> don't know whether this change has a positive effect.

Hello,

after running bind-9.3.2 for several days now with

#define ISC_MEM_USE_INTERNAL_MALLOC 1

compiled in, I believe I can say my problem was not version dependent
but malloc dependent.
One of our resolvers currently handles some 10.000 queries per second at
a load of 0.5 and an average CPU usage of 15-20%.

Kind regards
-- 
_____________________________
Marco Schumann