Re: [DMM] Comments to draft-hmm-dmm-5g-uplane-analysis-01

[Alexandre Petrescu]

Thank you very much for the clarification.

I understand now that GTP is a user-plane function run in the network.

- the UE (smartphone) never runs GTP.  I think this is worth putting in
  the draft and state it as such, if we so believe.

- it is strange for me to call it a User-Plane Function while the user
  has no impact on it.  I think this abbreviation deserves
  clarification.  We can suggest to 3GPP to change the name.

(terms like data-plane/control-plane (not user-plane) are already in 
widespread use, and could be improved).


Alex



Le 03/10/2018 à 18:29, Sridhar Bhaskaran a écrit :
 >>A UPF is any function that can be executed on user traffic in mobile 
network. Having said that however, I think the first UPFs that we will 
see will be in the form of SGW, PGW.


Even this is not accurate..

A UPF is a _*5G system core network*_ entity whereas the SGW-U and PGW-U 
are EPC core network entities. 5G core network has many differences from 
EPC. To avoid confusion it should be noted that 5G NR radio can be 
deployed with EPC as the core network and technically such deployments 
can also be called as 5G deployments. But an UPF has no role in such 
deployments.


Some of the user plane functionalities of an UPF are completely 
different from SGW-U or PGW-U.. Here are few differences (not exhaustive)


1. SGW-U and PGW-U --> one GTP-U tunnel per EPS bearer. No need of any 
QFI marking. UPF on the other hand has one GTP-U tunnel per PDU session. 
Different QoS flows within that PDU session are identified based on the 
QFI marking. So a UPF has to support QFI marking.
2. An UPF can support classification and encapsulation Ethernet frames 
whereas a SGW-U/PGW-U has no requirement to support classification and 
encapsulation of Ethernet frames.
3. An UPF supporting Ethernet PDU session type may support ARP proxying 
/ IPv6 ND Proxying whereas a SGW-U/PGW-U have no such requirement..
4. An UPF shall support RQI bit for reflective QoS. A PGW-U/SGW-U has no 
such requirement.


One may say that SGW-U / PGW-U also plays a role in user plane 
forwarding and hence why not call it a UPF?  For the lack of a better 
terminology, 3GPP has called the entity that performs the user plane 
functionalities and features within a _*5G core network*_ as UPF. Those 
functionalities are not exactly similar to EPC.


Thanks
Sridhar

On Tue, Oct 2, 2018 at 10:45 PM Arashmid Akhavain 
mailto:arashmid.akhav...@huawei.com>> wrote:


Not necessarily,

__ __

You are correct in that PGW is a UPF, but a UPF is by no means
limited to PGW. A UPF is any function that can be executed on user
traffic in mobile network. Having said that however, I think the
first UPFs that we will see will be in the form of SGW, PGW.

__ __

Arashmid

__ __

*From:*dmm [mailto:dmm-boun...@ietf.org
] *On Behalf Of *Behcet Sarikaya
*Sent:* Tuesday, October 02, 2018 11:44 AM
*To:* Shunsuke Homma mailto:homma.shuns...@lab.ntt.co.jp>>
*Cc:* dmm mailto:dmm@ietf.org>>
*Subject:* Re: [DMM] Comments to draft-hmm-dmm-5g-uplane-analysis-01

__ __

UPF is virtualized PGW, folks.

While PGW is fixed in location and possibly serving a large number
of UE which are geographically in the area, that part of the city
and that city itself, UPF can be deployed closer to the UE and thus
probably serving smaller number of UEs

__ __

Behcet

__ __

On Mon, Oct 1, 2018 at 10:47 PM Shunsuke Homma
mailto:homma.shuns...@lab.ntt.co.jp>>
wrote:

Thanks Dark for your explaining what is UPF instead of me.

Alex, brief definitions of UPF are described in the section
4.1.1.1 of
this draft. Also, you can find more details in 3GPP TS23.501.

Regards,

Shunsuke

On 2018/10/01 20:32, dirk.von-h...@telekom.de
 wrote:
 > Hi Alex,
 > and sorry for jumping into the discussion...
 >  From my and (AFAIK) 3GPPs understanding your smartphone is a
UE - sitting on the other side of RAN (gNB) - whereas a UPF
normally is seen as UP entry (and exit) of the 5G core (i.e.
handling all UP traffic in a true CP/UP split fashion).
 > Any other ideas on this? Can someone imagine any scenario
where UE implements UPF?
 > Thanks!
 > Best Regards
 > Dirk
 >
 > -Original Message-
 > From: dmm [mailto:dmm-boun...@ietf.org
] On Behalf Of Alexandre Petrescu
 > Sent: Montag, 1. Oktober 2018 13:22
 > To: dmm@ietf.org 
 > Subject: Re: [DMM] Comments to
draft-hmm-dmm-5g-uplane-analysis-01
 >
 >
 >
 > Le 01/10/2018 à 05:50, Shunsuke Homma a écrit :
 >> Hi all,
 >> # Sorry for my late response...
 

Re: [DMM] Comments to draft-hmm-dmm-5g-uplane-analysis-01

[Arashmid Akhavain]

Well put Sridhar… Thank you for the detailed clarification.

Speaking of QoS Flow Identifiers, I am assuming that mapping the QoS 
requirements from 5G core to the underlying transport nodes (routers, switches) 
and vice versa has not been changed. Is this correct Sridhar?

Arashmid

From: Sridhar Bhaskaran [mailto:sridhar.bhaska...@gmail.com]
Sent: 03 October 2018 12:29
To: Arashmid Akhavain 
Cc: sarik...@ieee.org; homma.shuns...@lab.ntt.co.jp; dmm@ietf.org
Subject: Re: [DMM] Comments to draft-hmm-dmm-5g-uplane-analysis-01

>>A UPF is any function that can be executed on user traffic in mobile network. 
>>Having said that however, I think the first UPFs that we will see will be in 
>>the form of SGW, PGW.

Even this is not accurate.

A UPF is a 5G system core network entity whereas the SGW-U and PGW-U are EPC 
core network entities. 5G core network has many differences from EPC. To avoid 
confusion it should be noted that 5G NR radio can be deployed with EPC as the 
core network and technically such deployments can also be called as 5G 
deployments. But an UPF has no role in such deployments.

Some of the user plane functionalities of an UPF are completely different from 
SGW-U or PGW-U. Here are few differences (not exhaustive)

1. SGW-U and PGW-U --> one GTP-U tunnel per EPS bearer. No need of any QFI 
marking. UPF on the other hand has one GTP-U tunnel per PDU session. Different 
QoS flows within that PDU session are identified based on the QFI marking. So a 
UPF has to support QFI marking.
2. An UPF can support classification and encapsulation Ethernet frames whereas 
a SGW-U/PGW-U has no requirement to support classification and encapsulation of 
Ethernet frames.
3. An UPF supporting Ethernet PDU session type may support ARP proxying / IPv6 
ND Proxying whereas a SGW-U/PGW-U have no such requirement.
4. An UPF shall support RQI bit for reflective QoS. A PGW-U/SGW-U has no such 
requirement.

One may say that SGW-U / PGW-U also plays a role in user plane forwarding and 
hence why not call it a UPF?  For the lack of a better terminology, 3GPP has 
called the entity that performs the user plane functionalities and features 
within a 5G core network as UPF. Those functionalities are not exactly similar 
to EPC.

Thanks
Sridhar

On Tue, Oct 2, 2018 at 10:45 PM Arashmid Akhavain 
mailto:arashmid.akhav...@huawei.com>> wrote:
Not necessarily,

You are correct in that PGW is a UPF, but a UPF is by no means limited to PGW. 
A UPF is any function that can be executed on user traffic in mobile network. 
Having said that however, I think the first UPFs that we will see will be in 
the form of SGW, PGW.

Arashmid

From: dmm [mailto:dmm-boun...@ietf.org] On Behalf 
Of Behcet Sarikaya
Sent: Tuesday, October 02, 2018 11:44 AM
To: Shunsuke Homma 
mailto:homma.shuns...@lab.ntt.co.jp>>
Cc: dmm mailto:dmm@ietf.org>>
Subject: Re: [DMM] Comments to draft-hmm-dmm-5g-uplane-analysis-01

UPF is virtualized PGW, folks.
While PGW is fixed in location and possibly serving a large number of UE which 
are geographically in the area, that part of the city and that city itself, UPF 
can be deployed closer to the UE and thus probably serving smaller number of UEs

Behcet

On Mon, Oct 1, 2018 at 10:47 PM Shunsuke Homma 
mailto:homma.shuns...@lab.ntt.co.jp>> wrote:
Thanks Dark for your explaining what is UPF instead of me.

Alex, brief definitions of UPF are described in the section 4.1.1.1 of
this draft. Also, you can find more details in 3GPP TS23.501.

Regards,

Shunsuke

On 2018/10/01 20:32, dirk.von-h...@telekom.de 
wrote:
> Hi Alex,
> and sorry for jumping into the discussion...
>  From my and (AFAIK) 3GPPs understanding your smartphone is a UE - sitting on 
> the other side of RAN (gNB) - whereas a UPF normally is seen as UP entry (and 
> exit) of the 5G core (i.e. handling all UP traffic in a true CP/UP split 
> fashion).
> Any other ideas on this? Can someone imagine any scenario where UE implements 
> UPF?
> Thanks!
> Best Regards
> Dirk
>
> -Original Message-
> From: dmm [mailto:dmm-boun...@ietf.org] On 
> Behalf Of Alexandre Petrescu
> Sent: Montag, 1. Oktober 2018 13:22
> To: dmm@ietf.org
> Subject: Re: [DMM] Comments to draft-hmm-dmm-5g-uplane-analysis-01
>
>
>
> Le 01/10/2018 à 05:50, Shunsuke Homma a écrit :
>> Hi all,
>> # Sorry for my late response...
>>
>> Thank you for your lively discussion. It is very helpful for
>> understanding points which need supplemental explanation and more
>> consideration.
>>
>> Following the discussion, we're planning to update the I-D for
>> covering the points below:
>>
>> - termination points of GTP-U
>> (RANs and UPFs terminate GTP-U in 5GS.)
>
> What is UPF?
>
> I understand UPF stands for User-Plane Function.
>
> Is my smartphone supposed to implement UPF?
>
> Alex
>
>> - setting QoS parameter of outer IP header
>> (Note that it's not just copy of inner