[Make-wifi-fast] [Bloat] Does 5g have the bloat problems of WiFi?

Dave Taht dave.taht at gmail.com
Thu Aug 1 19:30:45 EDT 2019

This is one of those philosophical questions that covers a lot of
ground under the 5g moniker.

0) This looks like a paid placement structured like an infomercial.
AT&T would dearly like X$/month from every device
connected to the network, selling microcells and carrying all your
traffic back to the cloud. The robot company is cute. The
story claiming this offloads the instore wifi - not so hot. Usually
the opposite claim is made.

1) For starters, it IS looking like AT&T etc, completely missed the
bufferbloat problem:


(I've got more links than this and have had a rather discouraging talk
with one of AT&T's managers )

Ericcson and a few other LTE related folk haven't, but they don't seem
to have products yet.

Perhaps they'll learn.

1a) For handsets there are decent cross layer techniques also.

2) But I think the thrust of the article re bandwidth and conflating
bufferbloat into it are *both* wrong. It's more a matter of how many
devices you can pack onto a given AP/microcell in a given area, which
is both a function of how good your (MU)-mimo is...
and how fast the mac can switch between them. Inter-device latency, not bloat.

And that's kind of anybody's guess at the moment. Including what 5g is
- 802.11d? y? a zillion other standards come under this umbrella. what
frequencies are you going to use? etc.

("bufferbloat" isn't a problem unless these robots are using huge
amounts of bandwidth, and I'd hope the were mostly just uploading
positional and other sensor data, rather than, say, a 360 degree
camera feed. However using the spectrum effeciently as the fq_codel
for wifi work did, certainly helps that)

(And were it me, I'd WIRE up the cash terminals, and sure as heck not
share the corp wifi with customers)

This being one of my bad days, I think the 5g/4g takeover is
inevitable given the billions being poured into it relatve to the
paltry investment into wifi. Outsourcing all the details as to how
your robot runs around, verses actually running a wire to your shop
and getting multiple APs right, few dead spots, is too hard to
maintain, handoffs too painful.

On my good days I point to intel doing a great job on their wifi chips
and closing down their LTE division, as an example that
LTE is actually far more difficult to make and make work than wifi is.
As well as nobody wants to have a sim card for ecery device they own
or outsource all their traffic to the cloud. WifI is still loved. It
is still a lot better than LTE in most coffee shops.

PS But vs the cost of the robot vs the cost of the wireless chipset,
I'd see 'em all equipped both ways and be letting the
customer decide.

On Thu, Aug 1, 2019 at 3:38 PM Kenneth Porter <shiva at sewingwitch.com> wrote:
> I just saw this article claiming that 5g is allowing brick-and-mortar
> automation providers to overcome limitations in WiFi, and I'm wondering if
> the technology is going to suffer from all the same problems previously
> seen in WiFi.
> <https://www.zdnet.com/article/googly-eyed-robots-actually-important-5g-edge-computing-benchmark/>
> (The "googly-eyed robots" in the title are man-sized robots that wander
> grocery stores to visually track inventory. Someone put googly eyes on them
> to keep them from scaring customers.)
> _______________________________________________
> Bloat mailing list
> Bloat at lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/bloat


Dave Täht
CTO, TekLibre, LLC
Tel: 1-831-205-9740

More information about the Make-wifi-fast mailing list