[afnog] Decreasing Access Time to Root Servers DNS by Running One on Loopback

[Mark Tinka]

On 17/Dec/15 01:42, Phil Regnauld wrote:

>
> 	I'd like to see data to back that up - not saying it's not
> 	likely, but I'd like to see real measurements :)

Measurements are always good :-).

Subjectively, we've seen faster resolution performance at conferences we
provide Internet access to when the local DHCP server is serving out our
network-wide resolvers rather than the resolver running on the local
resolver server physically at the conference. Fewer conference attendees
querying an on-site resolver than there are a bunch of customers
querying our network-wide resolvers, of course.

I'll consider measuring this at the next conference we serve.

That said, I suppose a less-populated resolver in Africa could perform
slightly worse than a less-populated resolver in the northern
hemisphere, accounting for the differences in latency to the root and
other zones, assuming the African resolver does not have (low latency)
access to a local root or local instance of various zones.

Mark.