ietf
[Top] [All Lists]

RE: [dnsop] Re: Root Anycast (fwd)

2004-10-02 04:55:23
Is there situation that multiple root servers installed behine multiple routers 
within one AS?
e.g. 
                                                               / ------ Router 
1 ---- Root Server-F-1
       AV8-----  ASBR-------Router-P  /
                                                             \
                                                               \  ----- Router 
2 ---- Root Server - F-2

If router-P enables PPLB, would there be some problem with TCP based DNS 
requests? 



Joe

-----Original Message-----
From: owner-dnsop(_at_)lists(_dot_)uoregon(_dot_)edu 
[mailto:owner-dnsop(_at_)lists(_dot_)uoregon(_dot_)edu] On Behalf Of Iljitsch 
van Beijnum
Sent: Friday, October 01, 2004 4:49 PM
To: Dean Anderson
Cc: dnsop(_at_)lists(_dot_)uoregon(_dot_)edu; ietf(_at_)ietf(_dot_)org
Subject: Re: [dnsop] Re: Root Anycast (fwd)


On 1-okt-04, at 2:48, Dean Anderson wrote:

Note though that it's *very* hard to create a setup where packets are 
delivered to different multicast instances, as it's hard to imagine 
how any real-world anycast setup could match the criteria in

Its quite easy for anycast: (real names used, but not real
relationships)

                Av8
              /   \
          sprint   att
              \   /
              F-root

If Av8 turns on PPLB, traffic to F-root will go through both sprint 
and att on a per-packet basis.

No, that won't happen because in order for BGP to install multiple 
paths in the routing table, the following conditions must be met:

"To be candidates for multipath, paths to the same destination need to 
have the following characteristics equal to the best path's 
characteristics:

[...]

•       One of the following:
        ◦       Neighboring AS or sub-AS (before the  eiBGP  Multipath feature 
was 
added)
        ◦       AS-PATH (after the  eiBGP  Multipath feature was  added)"

(From the URL I cited in my previous message.)

In this case either of the sub-conditions won't be met.

Per packet load balancing is inappropriate in cases where the entire 
path is different, as the limitations of the inferior path are imposed 
on the full set. BGP has enough knobs to distribute the load without 
this anyway. FIB-based load balancing is only appropriate when a small 
part of a single path is populated with more than one physical link. 
And in most of those cases, load balancing mechanisms that make sure 
one flow runs over one link are more desirable. Only in cases where 
this leads to a significantly unbalanced traffic distribution, per 
packet load balancing is a good idea. (But even then most people don't 
enable it as they are (too, IMO) afraid of packet reordering.)

.
dnsop resources:_____________________________________________________
web user interface: http://darkwing.uoregon.edu/~llynch/dnsop.html
mhonarc archive: http://darkwing.uoregon.edu/~llynch/dnsop/index.html


_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf