how much effect of memory on performance of bind?

[Jim Reid]

>>>>> "Cihan" == Cihan Subasi (Garanti Teknoloji) <CihanS at garanti.com.tr> writes:

    Cihan> 	I have 2 AIX, one is master one is slave but I am
    Cihan> having performance (quick response) problem on the slave
    Cihan> one, the only two differences are slave is bind8 and 64M
    Cihan> memory...Master is Bind4 and 384M memory....Any comment?

Get rid of BIND4 for starters. That code is long dead.

Based on the information you've provided, there's not much anyone can
say. A DNS lookup takes pretty much the same amount of time no matter
what version of BIND is involved. There can be differences depending
on how the servers and resolvers are configured and whether other
agents are involved in looking up a name or address, like
/etc/nsswitch.conf and things like NIS. Does the slow server forward
queries to another name server? Are there differences in resolver
configurations? Can you see which names the resolver(s) try when they
do a lookup? Is there a connectibity problem with the slave server?
Could lookups on one system be going through NIS before the DNS is
tried?

The amount of RAM  could make a difference if the zones are large
and there are memory-hungry applications running on the slower
server. But this would be showing itself in other ways: like processes
dying and fork() calls failing because the OS has run out of memory.