Bruce :)
[EMAIL PROTECTED] (Bruce Campbell) wrote:
> Recommendations for End-Sites:
...
> BGP-anycasted. Where possible, end-sites should not use
> PPLB for protocols which are likely to be BGP-anycasted,
> such as DNS.
Shouldn't one avoid this for all protocols? It adds quite some uncertainty
to the end site's routing, and most network operators will abstain anyway.
Think of debugging _any_ connectivity issue...
> If your listed, authoritative servers are utilising
> BGP-anycast, try to have at least one listed server for
> the zone which is unicast. This will continue to provide
> service to PPLB-using end-sites which have not turned off
> PPLB for specific protocols such as DNS.
I believe, almost nobody's crazy enough to go solely for anycast; alright,
alright, we all know the "bad guys", but apart from them, I've seen no-one
doing it.
There's still the issue of traffic "balancing" - consider a DNS setup with
one anycast NS and four unicast NS. Overall, _all anycast instances
together_ will receive about 1/5 to 1/4 (if lucky) of the entire query
traffic. That's quite a problem, and I'm not yet sure how to solve it
correctly.
But, using only anycast NS surely solves this economic problem.
> Remember that excessively long records will, quite likely,
> cause clients to flip from UDP to DNS. Try to avoid this
^^^
TCP
> on very popular zones by taking advantage of name
> compression or other tricks.
Elmar.
--
"Begehe nur nicht den Fehler, Meinung durch Sachverstand zu substituieren."
(PLemken, <[EMAIL PROTECTED]>)
--------------------------------------------------------------[ ELMI-RIPE ]---
.
dnsop resources:_____________________________________________________
web user interface: http://darkwing.uoregon.edu/~llynch/dnsop.html
mhonarc archive: http://darkwing.uoregon.edu/~llynch/dnsop/index.html
