5G roadblock: labor

Tue Dec 31 00:55:09 UTC 2019

On Mon, Dec 30, 2019 at 3:51 PM Christopher Morrow <morrowc.lists at gmail.com>
wrote:

> Oh good :) someone coaxed cameron out of the holiday keg :)
>

I can only take reading how others imagine it may work for so long

> On Mon, Dec 30, 2019 at 6:32 PM Ca By <cb.list6 at gmail.com> wrote:
> >
> >
> >
> > On Mon, Dec 30, 2019 at 2:41 PM Christopher Morrow <
> morrowc.lists at gmail.com> wrote:
> >>
> >> On Mon, Dec 30, 2019 at 4:11 PM Brian J. Murrell <brian at interlinx.bc.ca>
> wrote:
> >> >
> >> > On Mon, 2019-12-30 at 09:50 -0500, Shane Ronan wrote:
> >> > >
> >> > > Also, keep in mind that 10 years ago, you didn't know you would want
> >> > > or
> >> > > need 25mbits to your phone,
> >> >
> >> > Who needs 25mbits to their phone?
> >> >
> >>
> >> this is the wrong argument to make... or at the least it distracts
> >> from the conversation about: "Why 5g?" because everyone can come up
> >> with a reason for/against N mbps to Xthing. (I think this is sort of
> >> shane's point, actually)
> >>
> >> 4G/LTE:
> >>   o started the migration/consolidation of voice/video/data to a
> >> single bearer (well, IP anyway).
> >>   o moved the (ideally) IP endpoint closer to the tower base
> >>   o removed some latency, jitter, intermediaries between 'End-User'
> >> and "thing on the network"
> >>
> >> 5G:
> >>   o supposedly reduces latency 'more' (brings more of the IP
> >> connection and routing closer to the tower/radio?)
> >
> >
> > Latency to what? Latency between your handset and a front-end web server
> at Google or AWS is likely unchanged. Physic did not change for 5G.
> >
>
> good question... I think for any IP flow in previous deployments the
> point where my ip packet went from 'radio' to 'ip networking' could
> have been a fair distance away (super cell in 2g/3g worlds) from my,
> bending my IP path significantly from me to the thing I'm talking to.
> (introducing latency and other pokery from the carrier side swapping
> around from radio/3gpp/etc to "ip on ethernet").
>
> In the LTE world it's POSSIBLE that that transition could happen at
> the tower base (unlikely, but possible, theoretically). So, given some
> regional network and aggregation / etc my IP packet's path COULD be
> 'better'. That should enable better latency/jitter/etc. In practice
> the 3g ~300ms to send a packet from 'reston virginia' to 'ashburn
> virginia' has become ~20-40ms.
>
> Note, I'm not super interested in point-to-point measurements, but the
> general path being 'better' for user packets.
>

In order for mobility to work, there has to be a topology abstraction for
the notion of anchor point where the user always is. This anchor point in a
mobile network may have been 2 locations in the usa 15 years ago (all users
are anchored to 1 of these 2 places), but may be closer to 60 locations
now.  Ymmv depending on your carrier.  But, there are still only ~10 major
internet peering locations

Cell sites are normally a hub and spoke design in a metro area.   FB / GOOG
/ AWS only pick-up traffic from eyeball networks in  ~10 places in the usa.
Networks optimize for delivering of tonnage to those 10 places. None of
that fundamentally changes between LTE and NR. Cell sites aggregate in
buildings, those building connect to peering points.

Minimally, a 5g network is just hanging radios just like in LTE, and
backhauling those radios to hubs, just like LTE.  mmWave requires more
radios, low band less radios.

That said, things have improved in the last 15 years. All mobile traffic
for one carrier i know used to go to Seattle or Atlanta.  Which was
hilarious, since most people live in NY or CA.  Nowadays, generally,
packets from a handset start destination routing one hop (5-10ms metro-e)
up from the cell site... so  the packets don’t find themselves on a path of
indirection beyond your local metro area, and this is likely not a detour
along the path to internet peering.

Albuquerque packets will find their way to peerings in denver or dallas.
Birmingham packets will find their way to Peerings in Atlanta or Miami ...

> > Just random samples of what people post online....
> >
> > Vzw 5g 19 ms
> >
> > https://twitter.com/donnymac/status/1164491035503976448
>
> 19ms from 'georgetown' to <unknown> so I can't really tell what the
> uplift on a straight ping from (for example) georgetown university
> campus -> <thing> might be.
> either way... maybe it's 12-14 ms (since the test seems to talk about
> Annapolis which ought not be more than 3-4 ms from DC proper on fiber)
> that's not so bad really.
>
> > Att 5ge 34ms
> > https://twitter.com/joelouis77/status/1196651360185462784
> >
>
> yea.... no endpoints specified so: "testing that the internet is on fire"
> :(
>
> > Sprint , this guy shows 27ms on LTE vs 34ms on 5g
> > https://twitter.com/robpegoraro/status/1202705075535257600
>
> i'm guessing he means: "north arlington virginia" to "washington
> dc"... 34ms is 'long' :( much more uplift on that than I'd expect.
>
> >
> >
> >>
> >>   o simplifies management? (maybe?)
> >
> >
> > Hahahaha. No. Because 5g does not replace anything. It is yet another
> thing.
> >
>
> :) "long term, when you decom 3g and 4g for 5g! you know, when 6g
> arrives..." :)
> It's amazing to me that there's not a unified management system to
> offer network management across radio technologies? and some
> requirement from the carriers to push the vendors to provide a
> standards based interface to keep that management system in play long
> term? Maybe there is and it's running YANG/OpenConfig/etc ? Maybe it's
> silly to want that though because the radio world is 'so very
> different' from the plain-jim IP world? and/or there's enough
> difference between 3/4/5g that using a single management system just
> isn't practical?
>

There is talk of this... but these too are just more things. It is fun
reading all the press releases about open source stuff from at&t and then
trying to find the code and README file. Still looking.

> >> keeping speed out of the conversation, the footprint for MANY 5g
> >> deploymetns in the US (and elsewhere) is likely 'hundreds of feet
> >> circles', where 4G is 'miles'.
> >> (yes you can beam-form and make ovals and such...)
> >
> >
> > This is still a physics thing. Most purest will says 5G = new radio
> (NR). NR can run in any band. And, the distance is a function of the band.
> Tmobile is big on 600mhz NR, Sprint is big on 2500mhz NR and VZW has 28ghz
> NR.
> >
>
> oh sure, who gets the 'right' spectrum is going to drive what each
> carrier can 'do' with the 5g.
> but generally speaking, particular carriers aside... "5g is beneficial
> to users because?" (again, aside from speed increases).
>
> "Lower cost, because more people on less equipment and lower
> management costs" (pass that on to the users, right?)
> "more reach into places where cell coverage is spotty?"
> "new tech options over the network provided?"
>

Vendors are not interested in reducing costs to network operators, in
general. They may have replaced NPUs with x86 to reduce their own costs....

IMHO, operationalizing more spectrum into larger aggregation groups is the
thing NR gives over LTE.

Ask someone else, they may say talking cars and robot surgery and running
k8s on openstack.

> > There is an relation between the available spectrum bandwidth and the
> mhz. Meaning, there is only little 600mhz but there is a lot of 28ghz
> mmwave. That said, 600mhz can drive for miles, while 28ghz needs line of
> sight.
> >
> > Horses for courses. No silver bullet.  All the best mid-band spectrum ,
> balanced volume and propagation, got deployed in the 90s as pcs / gsm, and
> re-deployed in the 10s as umts and LTE.  Low band is great for penetration
> and coverage with a few cells, mmwave is great with line of sight...
> midband is the sweet spot in the middle.
> >
> > The future requires all tools available.
> >
> >>
> >>
> >> It'd be nice to see what benefits 5g really has for carriers and
> >> consumers/users... It looks, to me, like a bunch of the 5g hype is
> >> really: "uhm, we need to sell these carriers on the G++ ... spin up
> >> the hype machine about speed!" never mind the cost to deploy, range of
> >> deployment, changes in handset/radio gear / etc... more $ to the
> >> vendors!
> >>
> >> -chris
> >
> >
> > NR does operationalize more spectrum and allows bigger aggregate pools
> (like LACP) The new mmwave spectrum is the last to come to market because
> it’s value is limited in the general case.
> >
>
> thanks!
> -chris
>
> >>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.nanog.org/pipermail/nanog/attachments/20191230/9c194d29/attachment.html>