[TriLUG] What could be going on with my nameserver?
Aaron Joyner
aaron at joyner.ws
Tue Nov 1 18:22:14 EST 2005
Jon Carnes wrote:
>You really shouldn't be using forwarders anymore. And *don't* forward
>your DNS requests to a crappy low powered non-caching DNS forwarding app
>running on your Netgear router...
>
>You should have a file called "root.hint". your server will use this to
>populate itself with the current root Name servers. This lets your local
>Name server do direct lookup's for DNS requests.
>
>Here is an example setup for using the root.hint file...
>
Just to counter the absolute nature of the honorable Jon's statement a
bit, there are definitely some circumstances where a forwarder isn't a
bad idea. If you're on the other side of a slow link (ala a modem),
then a forwarder can shave literally seconds off every click you make in
a web browser. When the links involved are faster (your on some
broadband connection), you're still going to have a latency of 30-ish
milliseconds to the DNS server. So a forwarder (purely from a latency
standpoint) can shave maybe at most 50 to 100ms off every lookup you
make. That's not huge, but it's not negligible, but it's not the real
benefit of a forwarder. Consider the number and variety of queries sent
to your name server, vs the number and variety sent to (for example)
Time Warner's name servers? Virtually anything you're likely to request
is probably already cached on the TW name servers, because someone has
gone to that page recently. If you haven't gone to slashdot.org in the
last 7200 seconds (2 hours), you're going to be looking it up all over
again. On the other hand, it's really likely that some other geek on
broadband has gone to slashdot in the past couple hours, so you'll shave
your query time down from two queries (~120ms) to one query to a closer
server (network wise, at least, ~35 ms). Granted, we're still talking
in the sub-250ms arena, which isn't a big deal, but it's worth
mentioning the general benefit of forwarders.
Also, don't forget that it's the more net-efficient and thus
net-friendly way to run things, as you're reducing the load on the end
servers in favor of using a bigger more-local cache. The strongest
counter-argument is when you can't be dependent on the forwarders to
either a) reliably give you the right answer (security) or b) always
give you an answer (reliability). I'll leave those arguments to be made
by someone else. :)
Aaron S. Joyner
More information about the TriLUG
mailing list