On 6 aug 2009, at 19:00, Noel Chiappa wrote:
I see no particular issue with a network where some LAG-aware routers
do include the flow label in the hash and others don't.
Any time you have a network which is using hop-by-hop path selection
(i.e.
each node makes an independent decision on the next hop) and those
nodes are
not all using i) the same algorithm on ii) the same data, you can
have routing
loops.
Depends on whether you are using strong multipath routing or weak
multipath routing. (I hope there is better existing terminology for
this, but haven't found it so far.)
Consider:
A B
| |
+------+
| |
| |
| |
+------+
| |
C D
In the strong multipath routing case traffic from C to A can flow
directly and C-D-B-A. So far so good. But D's traffic can flow D-C-A
and D-B-A. This means that traffic towards A will flow over the C-D
link in BOTH directions. This can only happen when C and D can
determine for each packet whether it is on the C-D-B-A path or the D-C-
A path. So there must be something in the packet that determines this
and that is interpreted the same way by C and D.
In the weak multipath routing case this problem doesn't exist, because
packets towards a certain destination may only traverse a certain link
in one direction.
At this point, you won't be surprised to learn that existing multipath
routing techniques, including link aggregation load balancing of
various kinds, tend to use weak multipath routing. The exception is
OSPF ToS routing, which uses the type-of-service bits to keep things
straight. (Don't remember how the new and improved OSPF multitopology
routing does this...)
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
