Re: [Detnet-dp-dt] one question on S-DetNet-PEs

[Balázs Varga A <[e-mail]>]

Hi Loa,

Exactly. A,B and C are PE for their directly connected CEs, but P for pass-through traffic.

Let's see how "dense" DetNet flows will be populated in future networks. In networks

where only a small portion of the traffic require DetNet treatment we may see some

P nodes.

Cheers

Bala'zs



-----Original Message-----
From: Detnet-dp-dt [mailto:[e-mail]] On Behalf Of Loa Andersson
Sent: Thursday, March 09, 2017 12:48 PM
To: [e-mail]
Subject: Re: [Detnet-dp-dt] one question on S-DetNet-PEs



Bala'zs,



Oh, sure understood - it is just that when we invented the P and PE terminology it was not for nodes, but for functions. A PE were the group of functions needed in a node that interfaced customers. A P was the functions that were need by a node that was passing traffic inside a provider network,



The caveat was that PE an P functions, at least in the network that did run at that time, were often found on the same node, consider this.





           +-----+                     +-----+

+--+      |     |                     |     |     +--+

|CE|------|  A  |---------------------|  B  |-----|CE|

+--+      |     |                     |     |     +--+

           +-----+                     +-----+

              |                           |

              |         +------+          |

              |         |      |          |

              +---------|  C   |----------+

                        |      |

                        +------+

                            |

                            |

                          +---+

                          |CE |

                          +---+



If one view A, B and C as node are they P or PE nodes?



/Loa



On 2017-03-09 18:31, Balázs Varga A wrote:

> Hi,

>

> Correct. My comment was an additional argument for the tunnels.

> There are two scenarios mandating tunnels:

> 1, regular P nodes

> 2, DA-S-PE nodes acting as P node for some PWs

>

> Cheers

> Bala'zs

>

> -----Original Message-----

> From: Loa Andersson [mailto:[e-mail]]

> Sent: Thursday, March 09, 2017 11:11 AM

> To: Balázs Varga A <[e-mail]<mailto:[e-mail]>>;

> [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>

> Subject: Re: [Detnet-dp-dt] one question on S-DetNet-PEs

>

> Balázs,

>

> inline please.

>

> On 2017-03-09 17:02, Balázs Varga A wrote:

>> Hi,

>>

>> Generally agree, just two additional notes/comments:

>> - Mandatory tunnels:

>> They are needed anyway for regular P nodes. The DetNet role of

>> intermediate "P" nodes is limited to ensure congestion protection.

>> However P nodes can usually recognize only LSP labels and cannot

>> consider the whole label stack for flow recognition.  Therefore if

>> DetNet flows would not use tunnels P nodes may not distinguish them

>> from regular (non-DetNet) flows and cannot achieve congestion protection.

>

> While I agree to this, and it is another nail in the coffin of thw "we don't need", you are changing the scenario I draw. The point I tried to make was that DA-S-PE capable nodes in some scenarios might serve in a P role in some scenarios - the operators choice.

>

> /Loa

>

>

>>

>> - DA-S-PE needs also PW and FRER specific configuration:

>> As clarification to the step when "DetNet-PW-label is allocated by

>> the DA-S-PE for the DA-T-PE". You wrote:

>>    > the DetNet-PW-label is allocated by the DA-S-PE for the DA-T-PE, and

>>    >  as the LSP is set up an instruction is entered into the LFIB whether

>>    >  the DA-S-PE should do FRER or not.

>> I think we need more than a simple instruction "to do FRER or not".

>> The egress replication on the DA-S-PE (towards the next DA-S-PE(s) or

>> the terminating

>> DA-T-PE) may be DetNet-flow specific (how many member flows should be

>> created, which tunnel should be used by the egress member flows, etc.).

>>

>> I think we do not have signaling for that DetNet-flow and FRER specific configuration.

>> Do we intend to configure them in advance via management?

>>

>> Cheers

>> Bala'zs

>>

>> -----Original Message-----

>> From: Detnet-dp-dt [mailto:[e-mail]] On Behalf

>> Of Loa Andersson

>> Sent: Thursday, March 09, 2017 2:31 AM

>> To: [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>

>> Subject: Re: [Detnet-dp-dt] one question on S-DetNet-PEs

>>

>> Jouni, et.al.,

>>

>> On 2017-03-09 04:18, Jouni Korhonen wrote:

>>> Good guestion.

>>>

>>> I guess, if you want a specific node to be a S-DetNet-PE for some

>>> overlay and P for other, you could "tunnel" through it in P case.

>>

>> Yes - I think that is necessary. There is a bit more too it though.

>>> The

>>> currect draft still has text for "overlay labels" (i.e., L-labels)

>>> and I think those would work nicely for this case as you say below.

>>

>> OK - leave that in! There is a terminology issue here, the way we are doing things now, the L-label is the PSN tunnel in the PW architecture.

>>

>> The caveat is that PSN tunnel needsto be mandatory, unless you want to have a very complicated configuration for when a node is P for one overlay but DA-S-PE for another.

>>

>> Let me see if I got right

>>

>> - the PSN-tunnel (PHP'ed at the P node adjacent to the DA-S-PE) takes

>>    the DetNet PW to the DA-S-PE.

>>

>> - the DetNet-PW-label is allocated by the DA-S-PE for the DA-T-PE, and

>>    as the LSP is set up an instruction is entered into the LFIB whether

>>    the DA-S-PE should do FRER or not.

>>

>> - DA-T-PE establish a PSN-tunnel through which the DetNet PW is

>>    tunneled.

>>

>> - in the case of using a signaling protocol (since this is PWs I guess

>>    it by default is LDP) to establish the PW, no other node than the

>>    DA-S-PE, and the egress (DA-T-PE) sees the request for FRER.

>>

>> - for signaling the L-Label/PSN tunnel, RSVP-TE could be used, which

>>    means that the L-Label/PSN-tunnel also serves at BW container.

>>

>> Did I get that right?

>>

>> Yes I think it will work.

>>

>> /Loa

>>

>>>

>>> - Jouni

>>>

>>> 3/8/2017, 5:03 AM, Loa Andersson kirjoitti:

>>>>

>>>> Folks,

>>>>

>>>> There might be a simple answer, but I don't see it just now.

>>>>

>>>> Suppose we have a network, where we designate a certain number of

>>>> nodes as S-DetNet-PEs, to build a nice overlay DetNet.

>>>>

>>>> Assume that we also designate another set of nodes as S-DetNet-PEs

>>>> for another overlay DetNet.

>>>>

>>>> Also assume that some nodes that are S-DetNet-PE in one network are

>>>> P's in the other.

>>>>

>>>> If we signal that we have have a detnet-ms-pw going through a P

>>>> node that is capable of doing elimination/replication, how do we

>>>> stop the P-node from doing that?

>>>>

>>>> This was something that the T-Lables did for us.

>>>>

>>>> /Loa

>>

>



--





Loa Andersson                        email: [e-mail]<mailto:[e-mail]>

Senior MPLS Expert                          [e-mail]<mailto:[e-mail]>

Huawei Technologies (consultant)     phone: [phone]



_______________________________________________

Detnet-dp-dt mailing list

[e-mail]<mailto:[e-mail]>

https://www.ietf.org/mailman/listinfo/detnet-dp-dt