[HN Gopher] A novel channel contention mechanism for improving w...
___________________________________________________________________
A novel channel contention mechanism for improving wi-fi's
reliability
Author : belter
Score : 63 points
Date : 2024-10-13 13:58 UTC (8 hours ago)
(HTM) web link (arxiv.org)
(TXT) w3m dump (arxiv.org)
| mark254 wrote:
| Sounds like Token Ring...
| greatgib wrote:
| Regarding how commonly are used wifi and Bluetooth in our
| everyday life. I don't understand why we haven't more frequency
| bands available for it. It is more in the public interest than a
| lot of useless private initiatives like WiMAX that can easily get
| large frequency bands.
| PhilipRoman wrote:
| 6GHz is definitely a step in the right direction. Also filters
| out all the poor people with old devices so you can enjoy radio
| silence on your 320MHz channel ;)
| vlovich123 wrote:
| Temporary solution until within 5-10 years everyone has
| upgraded their routers and devices to sit on 6ghz, 5ghz, and
| 2.4ghz.
| dweekly wrote:
| Eh, also doesn't go as far, which forces higher deployment
| densities, which helps.
| vlovich123 wrote:
| The propagation differences between 5Ghz and 6Ghz are
| minimal compared with 2.4Ghz vs 5Ghz. In fact, given that
| there's a bunch of other protocol & HW improvements, it
| wouldn't be surprising to see identical 6Ghz and 5Ghz
| deployment density.
| the_mitsuhiko wrote:
| > I don't understand why we haven't more frequency bands
| available for it
|
| Do we really need more frequency? At this point it does not
| seem like the challenge for Wifi quality is actually the
| available frequency spectrum. In fact, I have a lot of devices
| still in this household that cannot connect to more than 2.4GHz
| and that is not a question of available spectrum but that
| supporting all those frequency bands apparently is too costly
| for some chips on the market.
| foota wrote:
| Perhaps if the bands were wider it would provide sufficient
| incentive for adoption, since there would be more channels
| available (and a wider band would presumably be easier than
| two separate bands that are further apart).
| wongarsu wrote:
| In any moderately dense city environment, apartment
| complexes, or high traffic areas you can absolutely feel the
| effects of a lack of frequencies.
|
| In American suburbia it might not matter much, but it's
| definitely an issue for the rest of us
| the_mitsuhiko wrote:
| I'm in a dense apartment block in Vienna. The walls are
| thick enough that I barely see two WiFi's. Automatic band
| selection means we end up in different frequencies.
| AlotOfReading wrote:
| Not everywhere in the world uses the same concrete walls.
| It's quite common to see dozens of access points trying
| to share the same three 20MHz channels available at
| 2.4GHz in places like the US or dense areas in Asia.
| leguminous wrote:
| I live in the US and I just counted 65 visible APs on
| 5GHz. The DFS channels aren't usable in this area, so
| almost all of those are trying to share the same 2 80MHz
| channels.
| tzs wrote:
| From what I've read a significant contributor to that
| congestion is routers using too much power. A router in the
| middle of an 800 ft^2 one bedroom apartment in a dense
| apartment complex doesn't need anywhere near as much power
| to cover the whole apartment as does a router at one end of
| an 2200 ft^2 4 bedroom ranch style house in a suburb.
|
| Out of the box most consumer WiFi routers will be
| configured for high power and most consumers won't even
| know that it is something they can change. Even if someone
| does know about it lowering it will probably make things
| worse for them unless their neighbors also lower their
| power.
|
| Maybe consumer WiFi routers should also include some kind
| of long range low bandwidth communication method, such as
| LoRa, to find and communicate with surrounding routers to
| build a map of the routers in a general area (not
| necessarily a spatial map [1]) and then agree on power
| levels and channel assignments to avoid interfering.
|
| [1] it would be a map of how they relate by WiFi signal
| strength rather than how they relate in space. So a two
| routers that are a couple meters apart but on opposite
| sides of a wall that nearly completely blocks WiFi radio
| frequencies would be far apart on the radio map despite
| being very close together spatially.
| Tarball10 wrote:
| Thanks to DFS restrictions in the US, there are effectively
| only two 80 MHz channels that can reliably be used in the
| 5GHz band.
|
| 2.4GHz only has three non-overlapping 20MHz channels, and
| those can only do ~286 Mbps throughput in the best case when
| using 802.11ax.
|
| The 6GHz band is finally allowing 14 non-overlapping 80 MHz
| channels and 7 160MHz channels, without any DFS restrictions
| (though some channels are lower-power/indoors only).
| vkdelta wrote:
| WiMAX has been dead for more than a decade now. Pretty rest of
| the spectrum is allocated to licensees for cellular/LTE/5G and
| other military applications.
| stefan_ wrote:
| Keep this in mind when someone tells you again what a beautiful
| demonstration of capitalism the "spectrum auctions" are. Most
| bits move over spectrum no one paid for.
| anotherhue wrote:
| I worked in this field for a while, and had my own novel
| mechanism for solving a problem.
|
| The major issue I had and I suspect this will have, is with
| devices that don't play by the rules. The unit economics mean the
| manufacturer is going to squeeze everything out, and if they can
| 'cheat' and claim higher numbers they will.
|
| Back ten years ago the issue was devices ignoring CTS frames but
| this feels similar.
| avidiax wrote:
| When I worked on WiFi at Microsoft, we discovered that Apple
| was cheating. Apple was implemented a random backoff timer to
| have less range than the spec, which would mean that Apple
| devices win most contentions.
|
| We decided not to copy this into Windows, since it would be a
| race to the bottom.
| ajb wrote:
| Wow. Do you (or anyone) know if they still do it? Was there
| any pushback on them?
| avidiax wrote:
| I don't know. It's possible that it was merely a bug, and
| maybe it was fixed at some point.
|
| But it's also not something that would cause noticeable
| issues on a network. Apple devices performing slightly
| better wouldn't be a big red flag. There are many reasons
| that this could be the case, from a better implementation,
| better testing, better HW (antennas), a different OS with a
| better TCP stack, etc.
| Salgat wrote:
| What's the legality of devices intentionally violating wifi
| standards and causing potential issues with other devices? Is
| this something the FCC could act on?
| linuxlizard wrote:
| There's no laws about it. It's mostly a "handshake
| agreement" enforced via the WiFi Alliance. You go through
| the WFA certification process to get the "WiFi Certified"
| sticker. https://www.wi-fi.org/certification
|
| Usually large vendors try to do the right thing.
| seagullz wrote:
| If it got tested and verified, wasn't their WiFi Alliance
| certification supposed to be revoked?
| avidiax wrote:
| I would guess that this aspect of the spec was not well
| tested. It is possible to write a test that causes 100 or
| 1000 collisions and plots the random distribution of the
| backoffs, but that is pretty complicated versus just
| checking that 1 collision had a backoff that was within
| range.
| orev wrote:
| Was it a small difference from the spec, like an off by one
| error might be, or was it an order of magnitude difference
| that might point to it being significant and intentional? (Of
| course dropped 0 could also be a bug). That's a fairly
| important detail for a claim like this.
___________________________________________________________________
(page generated 2024-10-13 22:00 UTC)