Re: [Detnet-dp-dt] a quick look at the replication/discarding case for detnet flows

[Jiangyuanlong <[e-mail]>]

Loa, please see my comments inline.
I also have a question on the availability requirements of the DetNet, do we need to prevent any single point of failure in the whole MPLS network?

Thanks
Yuanlong

-----Original Message-----
From: Loa Andersson [mailto:[e-mail]] 
Sent: Sunday, January 15, 2017 12:09 PM
To: Jiangyuanlong; [e-mail]
Subject: Re: [Detnet-dp-dt] a quick look at the replication/discarding case for detnet flows

Yuanlong,


On 2017-01-14 19:45, Jiangyuanlong wrote:
> Loa,
>
> Thank a lot for the slides, a very good start!

Tnx!

> Maybe we just need one layer of PW label,

Currently I dob't see how this would work, can you explain plz?
[YJ] I will prepare a slide to explain the details. It may take a couple of days.

> while redefine the MS-PW paradigm to support DetNet data plane.

This a very tricky approach, we would need to "tweak" the paradigm in a backwards compatibel way.
I tend to think that if we add replication and discarding at merege point this is a big enough "tweak".
[YJ] yes, we need to introduce a new data plane mechanism to support this function.

> The T-PEs as defined in the PWE3 can encapsulate/de-capsulate the 
> DetNet frames,

yes - the PW paradigm allows you to add whatever encapsulation we need.

  and generate the sequence number.

Yes the function is there to guarantee that packets are not re-ordred.

> The S-PEs usually can swap the PW label or terminate the PW 
> completely;

wouldn't terminating the PW make the S-PE into a T-PE, isn't at the core of the definition of S-PE and T-PE?
[YJ] Swapping is normal, and terminating is unnecessary. Though the S-PE may need to look into the PW control word to get the sequence number of the packet.

  we can extend their operations to support replication and filtering based on the sequence number.

I - think yes, that cam be done, but I don't see that filtering on the sequence number is enough, some context (e.g. label) is needed, right?
[YJ] Yes, some lattice PW connectivity such as introduced in draft-ietf-pals-mpls-tp-dual-homing-protection may be needed. These lattice PW can be dynamically established for the DetNet service.

> LSP FECs can work as in the traditional MPLS,

yes they will, but I thought that the consensus in the meeting were that we don't want non-detnet traffic sharing the transport of detnet, right?
[YJ] maybe I got it wrong, but my understanding is: the DetNet PW is exclusively used for DetNet services, though the MPLS network can still carry non-detnet services. Maybe the service providers have more to say on this topic:)

but if any new DetNet mechanism is needed in P nodes in the future, we can also add new FECs quite conveniently.

Yes it can be done, but I think we need from the start to avoid piggy- backing of non-detnet traffic.

/Loa


>
> Thanks,
> Yuanlong
>
>
>
> -----Original Message-----
> From: Detnet-dp-dt [mailto:[e-mail]] On Behalf Of 
> Loa Andersson
> Sent: Saturday, January 14, 2017 12:29 PM
> To: [e-mail]
> Subject: [Detnet-dp-dt] a quick look at the replication/discarding 
> case for detnet flows
>
> Folks,
>
> I took a quick look at the replication / discarding case we talked about at the meeting this week.
>
> Please remember that so far this is not a proposal, just a demonstration that there is at least one way of doing it.
>
> What is there can bee revised, improved, changed or discarded if we find something better. Take a look a send comments.
>
> /loa
>
> Warning: I have a grave case of dyslexia, so any odd or inventive spelling is pure accident.
>

-- 


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