Re: UDLR: Comments on draft-ietf-udlr-experiments-00

Achmad Husni Thamrin wrote:
> 
> > ---
> > Section  4.5.1. DR Election
> >
> > The use of the word "listener", should perhaps be "source" ???
> 
> "listener" is the intended word.
> 
> The DR has two tasks:
>  - send register messages to the RP
>  - originate Join/Prune messages to the upstream router toward the
>    RP/source if a listener exists on the network segment.

Ah... OK. (I think if you fix the scenarios this is clear!)

> 
> In my understanding of Scenario D, there should be no multicast listener
> on the UDL segment, correct?
> 
> The paragraph about DR election states that "if a multicast listener
> exists ...", which is not the case for the Scenario D.
> The last sentence of the paragraph is:
>    In the Scenario 5, DR election result is not important as all UDL
>    nodes are multicast routers.
> There is a typo there: "Scenario 5" should be "Scenario D".
> I guess I should change the text into "if a multicast listener exists as
> in the Scenario A and B..." to make it clear.
> 
> >
> > ----
> > PIM Join Override?
> >
> > Finally, does anyone have experience of a large PIM-routed network
> > based on UDLR links?
> 
> We do not have a large UDLR environment.
> I experimented with cisco 2621 routers running PIM-SM and found out that
> the cisco I used did not suppress Join message.
>
> Does anybody have similar experiences like this?
> 
> >
> > Do we get cases where there are an implosion of Join-Overide messages
> > being issued by receiver multicast routers when one router sends a
> > prune?
> 
> Let me make an illustration.
> Suppose that the unidirectional link contains 1000 PIM-SM routers and
> there are 10 multicast streams flowing on the link. Furthermore, suppose
> that each router joins all streams.
> If a router no longer needs to forward a stream, it sends a Prune
> message toward the upstream router. The other routers have to send
> overriding Join message to keep the stream flowing on the link.
> If the overriding Join suppression is disabled, then there would be 998
> overriding Join messages flowing on the link. If at almost the same time
> there is a Prune message for each stream, there would be 9980 overriding
> Join messages flowing on the link.
> 
> This is the worst-case scenario for 1000 routers and 10 streams.
> 
> If overriding Join suppression is enabled, the question is about the suppression
> effectiveness.
> This is an implementation issue, mainly regarding the probability
> distribution function of the random timer used to generate overriding
> Join messages. (research issue)
> 
> I think we can show (and solve) the cases for large flat networks using
> simulation .

Yes - I think what I meant to ask about was if one receive router prunes,
other routers MUST see the prune to know whether they still want the data.
If there are many members - do they all send a PIM-JOIN-OVERIDE towards
the feed to refresh the forwarding of the group that was attempted to be
pruned? This is suboptimal, to receieve a PIM-JOIN-OVERIDE from all
routers - some (most) should ideally be suppressed. This takes a round
trip delay... so if I udnerstand correctly from your comment,
there is some tuning issues here and a trade-off in leave latency v.
network capcity used for PIM-JOIN-OVERIDE.... Is that it?


Gorry Fairhurst

> 
> --
> Achmad Husni Thamrin <[email protected]>
> 
> _______________________________________________
> UDLR mailing list
> [email protected]
> http://www.udcast.com/mailman/listinfo/udlr