Re: [Detnet-dp-dt] detnet LSPs and PWs

[Jouni Korhonen <[e-mail]>]

>>>> 
>>>> Also the problem of comparing sequence numbers arriving over
>>>> different physical interface seems to be a eal problem.
>>> 
>>> I do not see this as an issue if separate port are attached to e.g., same switch soc.  How separate physical ports end up to the same PW instance is an implementation issue. There are non-trivial issues if the separate physical ports are in different blades or say in different switch socs.
>>> 
>> So you are saying, it is sometimes a problem, sometimes not?
> 
> *IF* you can channel the PW through the same processor, and that has fast enough access to the s/n registry, then this works, but it does so at a cost of scaling and reliability.

When we are talking real high speed implementations, I do not necessarily see a need for a GPP to be involved. Scaling I do not sign, reliability yes. If the node/blade goes down with the SN dbs, they are gone. But on the other hand, in that case the PWs in general are gone terminating the failing node and need to be re-established.

> The n/w operators will not be happy with this, and we probably need OAM to verify it is happening, but if the core facing egress T-PE ports that service 

I do not understand the OAM verifying comment.

> the LSP set that service a particular set of PWs all go through the same forwarder, then it can be made to work. However this is at a cost of availability, generality, scaling and opex.

Eventually a detnet flow that experienced replication always need to go through a single point (e.g. a S-PE or T-OE forwarder) in order to merge detnet flows and eliminate duplicates properly. One can have multiple T-PE egress nodes (for reliability for example) but then it will the detnet end station that needs to do the final merge and elimination.

- Jouni

> If that meets you requirements, I am happy, I am just doing due diligence here.
> 
> - Stewart