Re: [arch-d] ETSI launches new group on Non-IP Networking addressing 5G new services

[Uma Chunduri <[e-mail]>]

 Hi Toerless,

Your observations are right, of course
Quick replies in-line [Uma]:

On Wed, Apr 8, 2020 at 1:54 PM Toerless Eckert <[e-mail]> wrote:

> John,


> I think 3GPP 3/4/5/6G is not a fitting example to apply

the IP-over-foo approach.


> 4G/5G "core" networks effectively are overlays on top of of some

typically IP based transport network. Today the overlay nodes are mostly

software in edge-data-centers/pops. So its more naturally an

overlay.



[Uma]: Right. But those overlays served well and all IP/MPLS transformation
started there 10 years ago (from LTE days). It's only 2 bearers effectively
till now.  And those were simply not enough for some of the critical
services envisioned in 5G networks beyond eMBB (so called  "slice"). Hence
you will continue to see some of the efforts in this thread as in the
title  of this thread (while I know some of the big operators  which
support that I am not plugged in that effort).


> There is some liaison process whereby IETF technologies are pitching

their benefits to 3GPP if used an an underlay (i reall having

heard about this for SRv6 or LISP). Yet, 3GPP seems to prefer not

to really change anything fundamental in their overlay architecture

from what i've been told (only tangentially watching this).

"Only simple, minimal layering integration is good integration".



[Uma]: Lot of discussions couple of years ago in various side meetings. But
at the end those were not so fruitful.  Partly because there was a
misguided effort to remove their heavily deployed overlay.  Obviously,
collapsing underlay and overlay won't help. LISP is different and it can
potentially solve the control plane signaling and other issues in 3GPP
domain but it's not that backward compatible (yet). But I am sure, Dino and
other might be working in future releases ..


The more fundamental problem is that the mobile services in most

service providers are where the money is coming from, so now

they want to re-design the hardware underlay in their name, aka: the actual

physical network infrastructure besides DCs and radio towers.

And i think 3GPP has no clue what to do there exactly. But i

am sure that something strange will evolve.


[Uma]:  Yes; we all heard E2E slicing at some point in last few years?
There is a lot in that E2E including L2 and IP/MPLS. You can get only so
far moving the edge. Radio (with NR) is dis-aggregated  and  it doesn't
help even for small sub-set of use cases, which can be solved by
moving your DCs/Edge closure (cost, complexity and security issues).

But the bottom line is underlay (IETF) needs to evolve independently of
5G/B5G (at best this is one use case), given the earlier futile
discussions/efforts as you mentioned above. This would benefit vast number
of use cases with new media, non-eMBB slices and industrial internet
seeking beyond-best-effort which can co-exist along with BE. While I am not
claiming I know fully, but yes, some of these aspects could belong rightly
in this forum perhaps.


Cheers!
--
Uma C.





>