Re: [Ila] [DMM] [E] Re: Fwd: New Version Notification for draft-herbert-ila-mobile-00.txt

[Marco Liebsch <[e-mail]>]

Kalyani,

steering traffic on N6 in between a correspondent service (Data Network, DN) and a mobile's UPF,
and this is what I referred to, does not require N6/Gi-LAN functions, just routing policies in the transport network.
This can be policies at the N6 edges, means at the UPF side and the DN side, serving as ingress/egress for the
mobile's downlink and uplink traffic. Still these policies need information from the mobility control plane,
which is at least the locator (current UPF for that traffic).

But I understood that N6 is out of scope of your document, so it's fine.

Thanks,
marco

From: Bogineni, Kalyani [mailto:[e-mail]]
Sent: Donnerstag, 8. Februar 2018 13:34
To: Marco Liebsch; Sri Gundavelli (sgundave); Dino Farinacci
Cc: [e-mail]; dmm
Subject: RE: [DMM] [Ila] [E] Re: Fwd: New Version Notification for draft-herbert-ila-mobile-00.txt

Marco:

There are 2 kinds of non-mobility nodes/functions:

-          One set is what you are referring to as N6 based functions, also called Gi-LAN functions that could be
chained to PGW/UPF on Gi/SGi/N6.

-          The second set of functions are related to transport (IP/MPLS) over which mobility traffic traverses.

The first set needs some kind of policy/charging control and will need interaction to the services interface.
The second set possibly don't need policy/charging control.

We need to see if some of the protocol choices like SRv6 are trying to address the second set also.

Kalyani

From: Marco Liebsch [mailto:[e-mail]]
Sent: Thursday, February 8, 2018 5:03 AM
To: Marco Liebsch <[e-mail]<mailto:[e-mail]>>; Sri Gundavelli (sgundave) <[e-mail]<mailto:[e-mail]>>; Bogineni, Kalyani <[e-mail]<mailto:[e-mail]>>; Dino Farinacci <[e-mail]<mailto:[e-mail]>>
Cc: [e-mail]<mailto:[e-mail]>; dmm <[e-mail]<mailto:[e-mail]>>
Subject: RE: [DMM] [Ila] [E] Re: Fwd: New Version Notification for draft-herbert-ila-mobile-00.txt

..sorry, correction in my first sentence below:
"True, control plane impact on data plane can be on N3, N9 but also on N6."..

From: dmm [mailto:[e-mail]] On Behalf Of Marco Liebsch
Sent: Donnerstag, 8. Februar 2018 10:50
To: Sri Gundavelli (sgundave); Bogineni, Kalyani; Dino Farinacci
Cc: [e-mail]<mailto:[e-mail]>; dmm
Subject: Re: [DMM] [Ila] [E] Re: Fwd: New Version Notification for draft-herbert-ila-mobile-00.txt

Hi Kalyani,

My take on your latest feedback:
> I think you are talking about 2 features: one for mobility that needs the database; another
> for non-mobility state transfer between user plane nodes (not necessarily mobility nodes).

True, control plane impact on data plane can be on N3, N9 but also on N9. The latter is probably what you classify as non-mobility.
My point was to not break the N4 but rather look towards a reasonable and extensible protocol solution so that mapping
rules can be carried over it. In such case the mapping DB may be co-located with the SMF or external to the SMF. My main point
was to not make the control plane configuring the data plane through the service-based interfaces.

About the second case, I think it's interesting enough to include N6 where the mapping system would control
the data plane to steer traffic to the mobile's UPF(s). Think about decentralizing the UPFs and relocating a
UPF mid-session. The plain old but popular DMM scenario: How to enable IP address- and session continuity
when the anchor UPF changes.

Best regards,
marco


From: Sri Gundavelli (sgundave) [mailto:[e-mail]]
Sent: Mittwoch, 7. Februar 2018 18:23
To: Bogineni, Kalyani; Marco Liebsch; Dino Farinacci
Cc: [e-mail]<mailto:[e-mail]>; dmm
Subject: Re: [Ila] [DMM] [E] Re: Fwd: New Version Notification for draft-herbert-ila-mobile-00.txt

Hi Kalyani,

For all the approaches that we are talking (ILA, LISP, SRv6 ..etc), there are two nodes that's where the translation/tunneling is happening. In a generic sense, it could be a layer-2 termination point, a first-hop router, or a transit router. When seen from 3GPP lens, there is only UPF and the N4 interface that you talk about. But, there is N3 (eNB to UPF) and then there are also other nodes where there is an impact.  The translation/tunneling/steering may very well happen on some router connected to the service cloud/core (on N6),  or at some exit router where there is no 3GPP UPF function and there is no N4.


So, there are two key questions here:

1.) Is the UPF only node that is impacted here? Is the assumption that these protocols are doing some translation/tunneling only on UPF node? Or, we can introduce a two types of IP forwarding nodes, one collocated with UPF and another without UPF. I know how this discussion will go in 3GPP; they will insist and say they we will never recognize any other node other what they created.

2.) Is N4 the only interface to these two types of node variants. Or we will have N4' to these both types of nodes from some AF (which can interwork with the service bus), and we don't' touch N4.

Marco's point is to keep this generic and not make this very UPF specific, as it will be too restrictive.



Sri




From: "Bogineni, Kalyani" <[e-mail]<mailto:[e-mail]>>
Date: Tuesday, February 6, 2018 at 1:23 PM
To: Marco Liebsch <[e-mail]<mailto:[e-mail]>>, Sri Gundavelli <[e-mail]<mailto:[e-mail]>>, Dino Farinacci <[e-mail]<mailto:[e-mail]>>
Cc: "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>, dmm <[e-mail]<mailto:[e-mail]>>
Subject: RE: [Ila] [DMM] [E] Re: Fwd: New Version Notification for draft-herbert-ila-mobile-00.txt


Marco, Sri:



Here is the services based 5G architecture.



[cid:image001.png@01D3A107.A98253D0]



SMF is a control plane entity and talks to the User plane functions (UPF) through N4 interface as specified in 3GPP TS 29.244.



Here are two variants:



Option 1: Mapping DB talks to the UPFs using a variant of N4 possibly.



[cid:image002.png@01D3A107.A98253D0]



Option 2: Here there is no direct interface between Mapping Db and UPFs. UPFs also support APIs like the control plane functions.



[cid:image003.png@01D3A107.A98253D0]



The architecture is extensible and additional control plane or user plane functions can be added.



Is this what you had in mind?



Kalyani



-----Original Message-----
From: ila [mailto:[e-mail]] On Behalf Of Marco Liebsch
Sent: Tuesday, February 6, 2018 12:09 PM
To: Sri Gundavelli (sgundave) <[e-mail]<mailto:[e-mail]>>; Dino Farinacci <[e-mail]<mailto:[e-mail]>>
Cc: [e-mail]<mailto:[e-mail]>; dmm <[e-mail]<mailto:[e-mail]>>
Subject: Re: [Ila] [DMM] [E] Re: Fwd: New Version Notification for draft-herbert-ila-mobile-00.txt



It could be a nice option to keep the data plane specific control (the mapping DB you refer to) in the user plane and take a common N4 to update the mapping DB in case of mobility. But I think that clashes with the clear data plane / control plane separation in nextgen. And: there may be data plane solutions which don't come with a control plane / mapping system. For these the N4 needs to disseminate complete forwarding states (an more, e.g. for chargeable event monitoring, device dormancy support etc.) to all relevant data plane nodes, means the ones that hold a state for the mobile.



Btw, in terms of compatibility with nextgen, so far N4 is terminated only in few types of core data plane nodes with a dedicated role. We may investigate options to push forwarding and further policies from the (nextgen) control plane to other data plane nodes which don't terminate N4.



marco



-----Original Message-----

From: dmm [mailto:[e-mail]] On Behalf Of Sri Gundavelli (sgundave)

Sent: Dienstag, 6. Februar 2018 04:07

To: Dino Farinacci

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

Subject: Re: [DMM] [Ila] [E] Re: Fwd: New Version Notification for draft-herbert-ila-mobile-00.txt



> The UPF sends IP packets. The UPF is part of the NGC core, right? So

>the packets from the UPF get to a map-resolver and map-server via IP.

>It's pretty simple. At least it should be.



Sure, that LISP control plane packet is an IP packet. But, every message that is going between CP and UP will be named and numbered in 3GPP specs, and so said in my first mail that its probably a new interface specific to LISP.



With any of the IETF protocols, PMIPv6/LISP/ILA, it can be argued that these are IP packets. But, we should note that there is interworking needed with the 3GPP authentication infrastructure, and the protocol specific control plane. Note that these protocols are not doing MN identity establishment. May be I could be wrong here on the assumptions you have around LISP MN capabilities, but to me MN is a standard 3GPP UE with no special capabilities such as MIPv6/LISP MN stack.







Sri













On 2/5/18, 6:52 PM, "Dino Farinacci" <[e-mail]<mailto:[e-mail]>> wrote:



>> Sure, but I assume the mapping table/DB is some where else in some

>>central  location and not on the UPF?

>

>True.

>

>> The question is how does the UPF fetch that entry and if the

>>interface for  that query is built on some 3GPP interface, or its

>>internal to LISP with  no bearing on the access technology.

>

>The UPF sends IP packets. The UPF is part of the NGC core, right? So

>the packets from the UPF get to a map-resolver and map-server via IP.

>It's pretty simple. At least it should be.

>

>Dino

>

>>

>> Sri

>>

>>

>>

>> On 2/5/18, 6:42 PM, "Dino Farinacci" <[e-mail]<mailto:[e-mail]>> wrote:

>>

>>> I don't know what you mean. If you put the xTR function on an UPF,

>>> then by LISP spec definition, Map-Request, Map-Reply, and

>>> Map-Register functionality is part of the UPF.

>>>

>>> Dino

>>>

>>>> On Feb 5, 2018, at 5:27 PM, Sri Gundavelli (sgundave)

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

>>>>

>>>> I suspect there might be a need for a new interface.

>>>>

>>>> Assuming the LISP mapping system stays in the control plane, next

>>>>to  SMF/AMF, and the xTR functions on the UPF, there needs to be

>>>>probably a  new interface along the lines of the N4, for managing

>>>>the LISP MAP  operations (Reg/Req/Reply/Notify..).  But, off course

>>>>if the mapping  system stays in the user-plane, may be there is just

>>>>interworking with  the  3GPP authentication interfaces.

>>>>

>>>> Sri

>>>>

>>>>

>>>>

>>>>

>>>>

>>>>

>>>>

>>>>

>>>>

>>>>

>>>>

>>>>

>>>>

>>>> On 2/5/18, 5:15 PM, "Bogineni, Kalyani"

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

>>>>

>>>>> Dino:

>>>>>

>>>>> Please look at 3GPP TS 23.501 to understand the architecture of NGC.

>>>>>We

>>>>> tried to explain that in the White paper.

>>>>> TS 23.502 has the procedures for the NGC. TS 23.503 specifies the

>>>>>policy  and charging control framework for NGC.

>>>>> CT4 has a technical report on protocol aspects for NGC in TR 29.891.

>>>>>

>>>>> Your draft needs to describe how it fits in the 5G architecture,

>>>>>right  now it only addresses 4G.

>>>>>

>>>>> Kalyani

>>>>>

>>>>>

>>>>>

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

>>>>> From: ila [mailto:[e-mail]] On Behalf Of Dino

>>>>>Farinacci

>>>>> Sent: Monday, February 5, 2018 7:32 PM

>>>>> To: Bogineni, Kalyani <[e-mail]<mailto:[e-mail]>>

>>>>> Cc: Tom Herbert <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>; dmm

>>>>><[e-mail]<mailto:[e-mail]>>;  Sri Gundavelli (sgundave) <[e-mail]<mailto:[e-mail]>>

>>>>> Subject: Re: [Ila] [E] Re: [DMM] Fwd: New Version Notification for

>>>>>draft-herbert-ila-mobile-00.txt

>>>>>

>>>>>> On Feb 6, 2018, at 5:04 AM, Bogineni, Kalyani

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

>>>>>>

>>>>>> Dino:

>>>>>>

>>>>>> Can you add a section to show how this proposal would fit in 5G

>>>>>> architecture?

>>>>>

>>>>> Can you be more specific in what you¹d like to see in the new

>>>>>section?

>>>>>

>>>>> There are references throughout the draft where you see eNodeB and

>>>>>pGW  that UPF functionality could be at the same network mode

>>>>>location.

>>>>>

>>>>> Dino

>>>>> _______________________________________________

>>>>> ila mailing list

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

>>>>>

>>>>>

>>>>>https://urldefense.proofpoint.com/v2/url?u=https-3A__www.ietf.org_m

>>>>>ail

>>>>>ma

>>>>> n_

>>>>>

>>>>>

>>>>>listinfo_ila&d=DwIGaQ&c=udBTRvFvXC5Dhqg7UHpJlPps3mZ3LRxpb6__0PomBTQ

>>>>>&r=

>>>>>Id

>>>>> iS

>>>>>

>>>>>

>>>>>ODh8aDRjdCeGgd9MznLHMYKgKcs_YSwXBDiaofh47oilzaXYRYETcBynUdpT&m=zf1K

>>>>>fRu

>>>>>4n

>>>>> UF

>>>>>

>>>>>

>>>>>sUT8IJVExPygA_iAC-h4BErkY_CE2ugM&s=oLQOKLOAxuYtjVD_qWMbiQjkmP9acy6A

>>>>>u0X

>>>>>6l

>>>>> pS

>>>>> iBvg&e=

>>>>

>>>

>>

>



_______________________________________________

dmm mailing list

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

https://urldefense.proofpoint.com/v2/url?u=https-3A__www.ietf.org_mailman_listinfo_dmm&d=DwIFAw&c=udBTRvFvXC5Dhqg7UHpJlPps3mZ3LRxpb6__0PomBTQ&r=IdiSODh8aDRjdCeGgd9MznLHMYKgKcs_YSwXBDiaofh47oilzaXYRYETcBynUdpT&m=lWGVj8Jd11JyGVLcPLOSIxTZ-YHY3VbtfD1mi2uqhOY&s=-EIvAEYOQusoChy_iwtR4nMkaEqRKBTKTJ8GDjADuCk&e=



_______________________________________________

ila mailing list

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

https://urldefense.proofpoint.com/v2/url?u=https-3A__www.ietf.org_mailman_listinfo_ila&d=DwIFAw&c=udBTRvFvXC5Dhqg7UHpJlPps3mZ3LRxpb6__0PomBTQ&r=IdiSODh8aDRjdCeGgd9MznLHMYKgKcs_YSwXBDiaofh47oilzaXYRYETcBynUdpT&m=lWGVj8Jd11JyGVLcPLOSIxTZ-YHY3VbtfD1mi2uqhOY&s=cwX6UkOqq2vREiCvsQ7GPBXgKsinbkDmmckbpGwi73o&e=