No subject

authoritative server.  Nearly all of mine are for a caching only
server.  The overheads have been different in the past. 

To check the MP scaling of the stack, I'd probably try running some
netperf *_RR tests.  Since netperf does not even attempt to handle a
dropped UDP datagram, I'd use aggregate TCP_RR tests with the request
and response sized set to match DNS.  This will indeed differ from DNS
reality in that multiple sockets will be in use rather than the single
socket.  However, if the stack itself stops near to the transaction
rate...

Configuring the named to listen on more than one IP may help increase
SMP performance.

IIRC there was work in threads in the 2.6 line, so trying a reasonably
contemporary 2.6 environment might be interesting.  On the Itanium
system, profiling with q-tools from HP Labs - specifically
q-syscollect - might be worthwhile:

 http://www.hpl.hp.com/research/linux/q-tools/

rick jones
-- 
Wisdom Teeth are impacted, people are affected by the effects of events.
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to raj in cup.hp.com  but NOT BOTH...