Peter Siklosi a écrit :
Alexandru Petrescu wrote:
Zach Shelby wrote:
The issues is that some scalar metric representing the cost to
the best ER through a Router should be advertised in RAs. What
goes in the metric is determined by some routing algorithm but
should be an easily comparable scalar for nodes to use simply to
determine which router to use as a default without knowing
anything about routing.
Let me ask clarification about 'best router', in this example
topology:
------ RA1 ------- RA3 ------- |Host
|---+--|Router1|-----|Router3|\ ------ | ------- -------
\_______Internet | ------- /
+--|Router2|--------------/ RA2 -------
-what do you assume Host should choose as 'best router' - Router1,
Router2 or Router3? -did I assume correctly the subnet structure
above?
[...]
In my mind, the simplest model would be to represent the cost as hop
count and in this case router1 would advertise the cost 2, while
router2 would advertise the cost 1. The host would then select
router2 based on lowest cost, which would in this case be the
shortest route to edge router.
I agree. But remark I represented Router1 and 2 as devices with two
interfaces, and a different subnet for each interface. Not sure this is
the case for 6LoWPAN ND.
Indeed there seems to be a need for a Cost to be put in the RA but not
sure it applies well in the subnet structure here, because 6lowpan ND
considers sometimes mesh-under intertwinned with route-over... which is
confusing.
Were it for Router1 and 2 in 6lowpan ND to have two real interfaces, two
different subnets, then I'd agree with the use of Cost as you suggest.
Otherwise I'm not clear what a hop is at all (in multihop option metric).
Alex
In practical cases, this would be infinitely better than
selecting routers by random, but still not perfect.
A better model would be to take link quality into account. In this
case, the cost going through router1 is a function of the link
quality of the link to router1 and the cost from router1 to the edge
router. It would probably be best to define the cost as the sum of
the cost to router1 and the cost from router1 to the edge router and
we should require the link cost to be at least one. By this
procedure, it would still be possible to use just hop count, but also
possible to use something slightly more sophisticated.
Peter
_______________________________________________ 6lowpan mailing list
[email protected] https://www.ietf.org/mailman/listinfo/6lowpan
_______________________________________________
6lowpan mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lowpan
