Optimisation options

Hi my nickname lamoura.

Guys, help me please.

I have two servers :
1º - 8 cores, 16 gb ram
2º - 16 cores, 16 gb ram

What best values for the options below?

# some optimisation options.
server:
        # use all CPUs
        num-threads: <number of cores>
  
        # power of 2 close to num-threads
        msg-cache-slabs: <same>
        rrset-cache-slabs: <same>
        infra-cache-slabs: <same>
        key-cache-slabs: <same>

        # more cache memory, rrset=msg*2
        rrset-cache-size: 100m
        msg-cache-size: 50m

        # more outgoing connections
        # depends on number of cores: 1024/cores - 50
        outgoing-range: 950

        # Larger socket buffer. OS may need config.
        so-rcvbuf: 4m
        so-sndbuf: 4m

Zitat von Leandro Anjos Moura <lamoura@bsd.com.br>:

Hi my nickname lamoura.

Guys, help me please.

I have two servers :
1º - 8 cores, 16 gb ram
2º - 16 cores, 16 gb ram

What best values for the options below?

# some optimisation options.
server:
        # use all CPUs
        num-threads: <number of cores>

        # power of 2 close to num-threads
        msg-cache-slabs: <same>
        rrset-cache-slabs: <same>
        infra-cache-slabs: <same>
        key-cache-slabs: <same>

        # more cache memory, rrset=msg*2
        rrset-cache-size: 100m
        msg-cache-size: 50m

        # more outgoing connections
        # depends on number of cores: 1024/cores - 50
        outgoing-range: 950

        # Larger socket buffer. OS may need config.
        so-rcvbuf: 4m
        so-sndbuf: 4m

Hello

i guess you have already read http://www.unbound.net/documentation/howto_optimise.html. The question remain what your servers are used for. If you use them as caching resolver for a larger number of clients you should use a libevent enabled version of unbound and raise the outgoing-range and num-queries-per-thread.

Regards

Andreas

Tks.