[arch-d] Beyond Best Effort ?! Re: ETSI launches new group on Non-IP Networking addressing 5G new services

[Toerless Eckert <[e-mail]>]

This really goes to what i also think is one of the core architectural questions.

There is an important amount of DiffServ in many private networks,
including service provider video to the ome, that users might
consider "Internet". There are IntServ uses for criticial traffic
in Service providers and Enterprises (RSVP-TE), also for higher
resilience (reroute with resource guarantees). TSN has  a lot
of critical manufacturing and othher industrial use cases, and
those all want L3 support (DetNet). 

On the oter hand, we have in IETF not even come up with a better
definition of fairness than don't kill all competing traffic,
we have AFAIK no agreement of what level of aggressiveness of TCP
congestion control should be the fairness standard (maybe L4S will
help here a bit), and we are far from even acknowledging that we
should also have solutions for differentiated dynamic bandwidth,
even though we already have good technology proof points that it
can be done. (Stuck in Net Neutrality ...). See below.

We have recently made a bit of progress in IETF to better
support less than best effort, that could actually make
bits less expensive too.

Aka: I agree that from 50,000 feet you are right, but when you
come closer, you see that the picture is less binary,
but that what you say its true for "Best Effort Internet",
but that is not really sufficient, and its not true for many
private networks. And it is going to become more of an
issue, the more we want to apply Internet technologies
for more critical infrastructure functions IMHO. Or
even for more differentiated services experiences.

I for once think its not a good QoS architecture when
presidents have to call up NetFlix's CEO and ask him
to reduce his companies pressure on the networks, because
networks themselves can not distinguish this entertainment
traffic from more critical traffic. Do you think this
is the right QoS architecture for the Internet ? Are
we going to have an IANA registries for the FAX numbers
of those CEO's ? ;-)) (yeah, sorry, couldn't resist).

Cheers
    Toerless

On Thu, Apr 09, 2020 at 12:40:08PM -0400, Henning Schulzrinne wrote:
> Since we're talking slides, I'll throw in my take:
> 
> http://www.cs.columbia.edu/~hgs/papers/2019/2019%20Internet40%20IIT-RTC.pdf
> 
> Every single failed network architecture has promised QoS by making bits
> more expensive; every successful technology has lowered the total cost of
> deployment and operations and making bits cheaper.
> 
> On Thu, Apr 9, 2020 at 12:26 PM Bob Hinden <[e-mail]> wrote:
> 
> > I wonder if this desire for new non-ip networking architectures, is really
> > a replay of the what I call the packet switching vs. circuit switching
> > debate.   Somewhat more recently Ethernet vs. Token ring.  Basically, can
> > networking architectures provide deterministic service (like circuit
> > switching), or is the Internet best effort OK.
> >
> > It was very common to hear that voice or video couldn???t possibly work over
> > an Internet style packet switching network.   I haven???t heard that in a
> > while, or at least until recently after the 5G hype started gearing up.
> >
> > I attended the lunch speaker session on 5G at the Singapore IETF, slides
> > can be found:
> >
> >
> > https://datatracker.ietf.org/meeting/106/materials/slides-106-edu-sessa-5g-impact-on-networks-edge-cloud-and-slicing-00
> >
> > Seemed to me it was proposing (see slide 13) an end to end 5G
> > network/service.  Lots of words like "Guarantee transport SLAs for each
> > service???, ???Optimize network resources as load conditions change???, etc.   I
> > think arguments proposing a new non-ip networking architecture is back to
> > wanting a more deterministic networking (aka circuit switching).   I
> > suppose it also has something to do with wanting to sell services to very
> > large numbers of devices (vs. the current Internet model a subscription per
> > enterprise or home user).
> >
> > I didn???t understand why all of this couldn???t be built today with the
> > current set of IP wireless and wired technologies.   I was thinking of
> > coming to the mic and ask the speaker if he had heard about the Internet,
> > but decided it wouldn???t be constructive :-)
> >
> > Bob
> >
> >
> > _______________________________________________
> > Architecture-discuss mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/architecture-discuss
> >

> _______________________________________________
> Architecture-discuss mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/architecture-discuss


-- 
---
[e-mail]