Chaining forwarders

Kevin Darcy kcd at daimlerchrysler.com
Tue May 2 00:02:16 UTC 2000


Ed Sawicki wrote:

> There is a note on page 246 of the DNS & BIND book (Third
> Edition) that cautions against chaining forwarders. The only
> reason I see mentioned in the book for this is making the
> configuration too difficult. Is this the only reason?
>
> I see benefit in chaining forwarders for certain network
> topologies.

Certain *broken* network topologies, you mean? To me, forwarding is a
Necessary Evil to get around network connectivity issues, i.e. server
A has no connectivity, or lousy connectivity, to network B, therefore
forwards to server C, which has better connectivity to that network.
Given this attitude, the only justification I'd see for chained
forwarders would be multiple levels of disconnectivity or bad
connectivity. At that point, I'd look at fixing the network topology
before I'd chain the forwarders.

I'll note that the recent enhancements to BIND to use RTT in making
forwarding choices improves the robustness of the forwarding mechanism.
But by its very nature forwarding is still not as scalable or optimizing
as iterative resolution.


- Kevin




More information about the bind-users mailing list