Re: [v6ops] Question on multi-homed nodes and address/route selection

[Tim Chown <[e-mail]>]

On 2 Mar 2017, at 16:28, Eric Vyncke (evyncke) <[e-mail]<mailto:[e-mail]>> wrote:

Simon,

You nail the problem that the draft https://tools.ietf.org/html/draft-bruneau-pvd-00 is trying to address ;-)

Comments are welcome of course (I will send a separate email for those comments in a few minutes).

The other document I was refering to is now expired: https://tools.ietf.org/html/draft-ietf-6man-multi-homed-host-10

I assumed you meant https://tools.ietf.org/html/draft-bowbakova-rtgwg-enterprise-pa-multihoming-01 :)

But the IETF has been kicking this particular football around for at least 15 years now, since before the multi6 WG was born and died… looking at https://datatracker.ietf.org/wg/multi6/documents/ is rather nostalgic!

Tim

-éric


On 02/03/17 14:45, "v6ops on behalf of Simon Hobson" <[e-mail]<mailto:[e-mail]> on behalf of [e-mail]<mailto:[e-mail]>> wrote:

   Eric Vyncke (evyncke) <[e-mail]<mailto:[e-mail]>> wrote:

How timely :-)

Together with other folks, we are submitting an I-D in the  coming hours about Provisioning Domains (PvD) which addresses your use case of multi-homing in a 'complex case'.

   I look forward to looking at it.

AFAIK, there is another I-D (or even RFC) that recommends hosts to select the default route next-hop based on the source prefix used.

   That doesn't really address the issue. It's logical that for whichever address the node picks to use as a source address for an outbound connection, it should use an appropriate outbound route - I would guess, logically a router that advertised the matching prefix as one it routes for.

   But you have to take a step back and look at how that source address is picked - because in effect, source address selection for an outbound connection is intrinsically linked to route selection. Ie, if the node/service picks address A then the packet must be routed via connection X, if it picks source address B then the packet must be routed via connection Y.

   So it's no good if (just picking an example) I've got a connection that is good for bulk transfers (cheap, maybe slower and/or higher latency/jitter) and want to route my outbound mail via that connection, thus leaving the higher cost but better connection for interactive use (eg browsing) - if the mail service then uses a source address meaning that the packets have to be routed via the latter connection.
   I could force the mail service to bind to just one address - but that then means more local config - and it breaks if the latter connection goes down and I need it to failover to the other one for the duration of the outage.


   _______________________________________________
   v6ops mailing list
   [e-mail]<mailto:[e-mail]>
   https://www.ietf.org/mailman/listinfo/v6ops


_______________________________________________
v6ops mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/v6ops