RE: [PWE3] PWE3 new WG items? PW QoS

["Busschbach, Peter B (Peter)" <busschbach at lucent.com>]

Peter> Assume one wants to set up a multi-segment from A via B and 
> C to D to carry ATM traffic. Suppose that the segment between 
> B and C is set up independently from the other two segments. 
> Then it is possible that B indicates to C that it accepts a 
> maximum number of ATM cells per frame equal to N1. 
> Subsequently, B could learn from A that A accepts a maximum 
> number of ATM cells per frame equal to N2. If N2 is smaller 
> than N1, there is a problem. B might have to repackage frames 
> it receives from C to make sure it doesn't exceed N2 cells in 
> the frames it forwards to A.
> 
> Based on this example, it seems obvious to me that B has to 
> receive a PW setup message from A before it can forward it to 
> C and C has to receive a setup message from B before it can 
> forward it to D. In other words, there is a clear order.
> 
Wei> Hi Peter,
> 
> Agree with your observation.  Even though the ordered setup 
> is a sensible 
> approach and makes life easier to establish a multi-segment 
> pseudowire, it's 
> still up to the stitching nodes, B and C in this case, to 
> decide what the 
> aggragated events should be to trigger its signaling.  For 
> example, they need 
> other resources to get ready before start signaling and they can wait 
> indefinitely before start signaling.
> 

Wei,

This is perhaps another example where it makes sense to describe the problem to be solved before discussing solutions.

In my view of the world, if you need a multi-segment PW from A via B and C to D, you can only achieve your goal if all four nodes are ready. I don't see why it makes sense to set up the segment between B and C if A or D are not ready yet.

I could even go further: I would argue that node B doesn't even know that it is in the path from A to D, until it receives a setup message from A. To protect against network failures, node A should be able to decide on the best route to destination D and node B might not lie on that route.

But again, let's formulate the problem first.

Peter

 
> Therefore, the ordered setup seems to be more of an 
> implementation specific 
> issue for the stitching nodes rather than an item to be standardized.
> 
> ---Wei
> 
> > 
> > Peter 
> > 
> > _______________________________________________
> > pwe3 mailing list
> > pwe3@ietf.org
> > https://www1.ietf.org/mailman/listinfo/pwe3
> 
> 

_______________________________________________
pwe3 mailing list
pwe3@ietf.org
https://www1.ietf.org/mailman/listinfo/pwe3