Re: [Detnet-dp-dt] Using MS-PW concept for the d-pw

[Jiangyuanlong <[e-mail]>]

Loa, 

-----Original Message-----
From: Detnet-dp-dt [mailto:[e-mail]] On Behalf Of Loa Andersson
Sent: Monday, February 27, 2017 6:55 PM
To: [e-mail]
Subject: Re: [Detnet-dp-dt] Using MS-PW concept for the d-pw

Yuanlong,



On 2017-02-27 16:41, Jiangyuanlong wrote:
> Jouni,
>
> I must admit that I am confused by the use of "T-label" and "L-label".
This is from one point of correct, it is labels in the same label stack.
Naming the outermost label T-label is a long standing convention.
Naming the label below the T-label L-labels is also somewhat of a convention.
But, whatever it calls, I mean the label that stands for the MPLS path from T-PE to the S-PE.
Slightly incorrect, the LSP takes from the T.DetNet-PE to a S-DettNet-PE (I think we need to agree on terminology to communcate).
The outer label only takes you one hop.
[JY] Thanks much for the clarification. Then for all options, we will need this L-label, right?

> Please note that the LSP (Label Switching Path) from T-PE to the S-PE 
> does not require that the same label value be used for each node on 
> the path,
this is correct, the requirement is the opposite, the label changes every hop.
but may change hop by hop.
As I said - it will change at every hop. One reason (not the only) to have a label below the outer label in the stack is that the receiving node might have reason to know where the packet came from, the T-label does not give that info, the L-label does.
> For example, S-PE node (B) receives label 10 from node 2,
We normally don't use special purpose labels in examples
[YJ] Good to know that;)

while the label sent by node (A) is 20,

The label operations are correct (hop-by-hop) node (B) knows that the packets originated from node (A) if these two labels belong to the same LSP.

Let us say that you have used LDP to set up the LSP, when B receives the label you call 10, exactly how does it know that the LSP originated from A?

> Since LSP can be an hierarchy itself (i.e., LSP over another LSP tunnel), I suggest not to use both terms to simplify the problem space.

Well, I guess you are about two decades to late for that terminology reform. We have lived with it almost all the time MPLS has been around, it is not a coincidence that e.g. RFC 3209 was called  RSVP-TE: 
Extensions to RSVP for LSP Tunnels (draft-ietf-mpls-rsvp-lsp-tunnel).

[YJ] We need to setup two paths from T-PE1 to T-PE2 (not sharing fate if possible), traversing S-PE1 and S-PE2. I think RSVP-TE would be more likely to serve our purpose than LDP, it also provides bandwidth reservation and QoS guarantee.

Thanks again,
Yuanlong

/Loa
>
> Thanks,
> Yuanlong
>
> -----Original Message-----
> From: Jouni Korhonen [mailto:[e-mail]]
> Sent: Sunday, February 26, 2017 12:37 PM
> To: Jiangyuanlong
> Cc: [e-mail]
> Subject: Re: [Detnet-dp-dt] Using MS-PW concept for the d-pw
>
> Yuanlong,
>
>
>> On 25 Feb 2017, at 02:02, Jiangyuanlong <[e-mail]> wrote:
>>
>> Loa,
>>
>> -----Original Message-----
>> From: Loa Andersson [mailto:[e-mail]]
>> Sent: Saturday, February 25, 2017 3:34 PM
>> To: Jiangyuanlong; Jouni Korhonen
>> Cc: [e-mail]
>> Subject: Re: [Detnet-dp-dt] Using MS-PW concept for the d-pw
>>
>> Yuanlong,
>>
>> The L-laabel is ther to make the the intermediate not (S-PE) know what to do whit the packet, the d-pw label was not allocated by the S-PE, so it does not ahve any knowledge what it means.
>> [YJ] d-pw can be allocated by the S-PE in MS-PW, just as you would like to allocate the L-labels. Very similarly, I think the same T-LDP protocol can be used.
>> [YJ] As I said in the previous email, using PW to trigger FRER will be cleaner compared with using L-label since CW is inspected.
>
> I think you are still getting it wrong what was intended with L-labels. They were specifically thought in the context of detnet global d-pw labels.
>
> L-labels connected MS-PW nodes over the network topology. All FRER “triggering” etc is still and has always been tied to the d-pw. Within a S- or T-PE you need to pop the L-label and then inspect the top of stack d-pw label..
>
>                                                  within
>     LSR                         |------------- S- or T-PE -------------|
>                     PHP                            POP
>                   -------->                     -------->
> +------------+                  +------------+            +------------+
> |  T-labels  |                  |  L-label   |            | d-pw label |
> +------------+                  +------------+            +------------+
> |  L-label   |                  | d-pw label |            |            |
> +------------+                  +------------+            |  Payload   |
> | d-pw label |                  |            |            |            |
> +------------+                  |  Payload   |            +------------+
> |            |                  |            |
> |  Payload   |                  +------------+
> |            |
> +------------+
>
>
> This allowed also bypassing S-PE easily for some L-labels.. instead of an L-label pop one would do a swap and just forward after that.
>
>
> - Jouni
>
>
>> If you let the S-PEs allocate and swap d-pw's, the next S-PE or a T-PE can't coordinate for the same packet coming in on from tow different nodes.
>> [YJ] It's like the 1+1 PW protection case, though the operations of elimination and replication in the S-PE and the T-PE need to be specified.
>> [YJ]Take VPLS as an example, several PWs can be directed into the same VSI in a PE and PW packets are processed there (for detnet, the processing is FRER now).
>>
>> But I feel like we are going in circles, can we agree on the corner stones first?
>>
>> Do we want all possible/conceivable control mechanism be within scope?
>> [YJ] maybe we can take LDP as a first step. It seems the difficulty is how to decide the S-PEs for a detnet flow (a routing protocol may be needed for automatic selection).
>> [YJ] if all T-PEs and S-PEs are determined for a detnet flow, it is quite easy to set up PW segments and LSPs respectively with the help of LDP protocol.
>> Cheers,
>> Yuanlong
>>
>> /Loa
>>
>>
>> On 2017-02-25 15:18, Jiangyuanlong wrote:
>>> I agreed to alternative 2. L-label is not needed, S-PE must look into the PW label (further, extract sequence number in the CW) of a packet, and all FRER semantics can be coupled with the PW label.
>>>
>>> -----Original Message-----
>>> From: Jouni Korhonen [mailto:[e-mail]]
>>> Sent: Saturday, February 25, 2017 3:07 PM
>>> To: Jiangyuanlong
>>> Cc: [e-mail]; Loa Andersson
>>> Subject: Re: [Detnet-dp-dt] Using MS-PW concept for the d-pw
>>>
>>> Hi,
>>>
>>>> [YJ] I regard the L-Labels and T-labels are the same LSP layer. If we look into the full MPLS label stack of a packet in a PW, normally there is an LSP label at the top (unless PHP is enabled for the last hop).
>>>> Not sure what is the L-Label in your picture, is it different from LSP label?
>>>
>>> L-labels have been so far in the discussion between MS-PW PEs.
>>> T-labels are between any LSR. L-labels are not PHPed i.e., even if 
>>> PHP is enable the L-Label stays and the label above it gets popped 
>>> (that we have been referring as T-Labels). T- and L-labels are just 
>>> a naming convention. IF you don’t have “between MS-PW PEs” semantic 
>>> associated with the L-Label, it is the same as T-Label ;)
>>>
>>>
>>>
>>> case L-labels are present.. (alternative 1)
>>>
>>>                     PHP
>>>                   -------->
>>> +------------+                  +------------+
>>> |  T-labels  |                  |  L-label   |
>>> +------------+                  +------------+
>>> |  L-label   |                  | d-pw label |
>>> +------------+                  +------------+
>>> | d-pw label |                  |            |
>>> +------------+                  |  Payload   |
>>> |            |                  |            |
>>> |  Payload   |                  +------------+
>>> |            |
>>> +------------+
>>>
>>> case no L-labels.. (alternative 2.. and also alternative 3 if you 
>>> think T- and L-labels are the same)
>>>
>>>                     PHP
>>>                   -------->
>>> +------------+                  +------------+
>>> |  T-labels  |                  | d-pw label |
>>> +------------+                  +------------+
>>> | d-pw label |                  |            |
>>> +------------+                  |  Payload   |
>>> |            |                  |            |
>>> |  Payload   |                  +------------+
>>> |            |
>>> +------------+
>>>
>>>
>>>
>>>
>>>
>>> - Jouni
>>>
>>
>> --
>>
>>
>> Loa Andersson                        email: [e-mail]
>> Senior MPLS Expert                          [e-mail]
>> Huawei Technologies (consultant)     phone: [phone]
>> _______________________________________________
>> Detnet-dp-dt mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/detnet-dp-dt
>
> _______________________________________________
> Detnet-dp-dt mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/detnet-dp-dt
>

-- 


Loa Andersson                        email: [e-mail]
Senior MPLS Expert                          [e-mail]
Huawei Technologies (consultant)     phone: [phone]

_______________________________________________
Detnet-dp-dt mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/detnet-dp-dt