Eduardo Júnior wrote:
Hi,
I read about Bind which it works as follow (in general words):
query -> named -> named.conf -> files zones
Is it possible create many named.conf, one per thread with the
objetive of to reduce queue's lenght of querys.
For example:
named.a-e.conf
named.f-m.conf
named.n-z.conf
Some query to about.com <http://about.com> would be treated by
named.a-e.conf which would read the file zone about.com <http://about.com>
named already dispatches lookups to separate threads, according to load,
how do you think you'd be able to do better with an "alphabetic"
dispatch algorithm, rather than one based on load?
Note that answering DNS queries from authoritative data -- what you seem
to be describing -- is about the most lightweight possible thing that
named does. The heavy lifting comes into play when named tries to
resolve recursive queries from non-authoritative data; then it needs to
follow the iterative resolution process, working its way down the
delegation hierarchy, dealing with slow/unavailable nameservers,
misconfigured delegations, etc. And this doesn't lend itself to any kind
of "alphabetic" optimization: imagine, if you will, aaa.com being
delegated to mmm.com nameservers, which then resolve www.aaa.com as a
CNAME to www.zzz.com, which is then delegated to eee.com nameservers,
etc. etc. etc. Are you going to hand off all of those different parts of
the resolution process, to different threads? That's an awful lot of
"thrashing", and likely to erase any benefit you hope to gain.
- Kevin
_______________________________________________
bind-users mailing list
bind-users@lists.isc.org
https://lists.isc.org/mailman/listinfo/bind-users
