I am using BIND version 9.3 on HP and noticed that the forwarder
behavior has improved as compared to the previous version
( BIND 9.2 ).

When the first forwarder in the list is down, named remembers the
best forwarder and forwards all the queries to the best forwarder.
named
also modified the srtt values of all the forwarders after every query .

My observation is that named uses SRTT to identify the best
forwarder - lower the value higher the preference. We can use
the 'rndc dumpdb' to dump the named contents and the dumps
are in named_dump.db.

Question:

1. What is the algorithm used? Is it similar to the SRTT in TCP?
2. Where can I find more details on this?

Any information regarding this is greatly appreciated.

thanks

regards

Blr