[dns-operations] TCP anycast was Re: DNS query logging

Edward Lewis Ed.Lewis at neustar.biz
Fri Mar 18 19:27:06 UTC 2011


At 14:03 -0700 3/17/11, Rick Jones wrote:
>If there is jitter in the RTT with an anycast NTP server it suggests the
>"routing" is changing right?  At least once over an interval of either
>64, 128, 256... 1024 seconds.  So, while NTP is more time-sensitive than
>DNS, its sensitivities underscore that generic TCP with anycast isn't a
>slam dunk in the positive.  At least that is my current interpretation.

At 15:10 -0700 3/17/11, Rick Jones wrote:
>Beyond that to the server via unicast that is... just in case that
>wasn't obvious.

The routing in anycast and unicast is the same.  In fact, anycast is 
"isomorphic" to a single site that can be reached via multiple paths 
(multihomed).

What I was saying is that perhaps NTP is more sensitive to jitter, 
thus wouldn't appreciate fallback to another path.  There are times 
when a hard failure is more useful than a degradation.

As far as the rubber banding being an indication of routing, I can't 
say.  Rubber banding used to be more common when we had links with 
high drop rates.  In NTP's case, rubber banding would be difficult to 
overcome and probably would rather just drop the source.

At 14:03 -0700 3/17/11, Rick Jones wrote:
>Does a shift manifest itself as a failed query, and how is that
>generally dealt with?  For example will it cause an EAI_AGAIN or a
>different error?

Perhaps.  But then again maybe not.  TCP is a relation between two 
end points, in between it is IP.  If the path changes out below a TCP 
session and the routing converges before a timeout happens, TCP 
wouldn't care.  However, it's been decades before I had deep enough 
knowledge to give any specifics.

With UDP, any sneeze may lose the packet.  But even if it whirlpools 
a bit before getting home, so long as it's in the time window, it's 
okay.

...There was a time when jitter could be the result of many things, 
maybe today it's reduced to route changes.  I can't say.  (Although I 
do know flaky wireless still abounds.)

-- 
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Edward Lewis
NeuStar                    You can leave a voice message at +1-571-434-5468

Me to infant son: "Waah! Waah! Is that all you can say?  Waah?"
Son: "Waah!"



More information about the dns-operations mailing list