[HN Gopher] Wi-Fi 7 Signals the Industry's New Priority: Stability
___________________________________________________________________
Wi-Fi 7 Signals the Industry's New Priority: Stability
Author : CharlesW
Score : 39 points
Date : 2023-12-29 21:05 UTC (1 hours ago)
(HTM) web link (spectrum.ieee.org)
(TXT) w3m dump (spectrum.ieee.org)
| Firerouge wrote:
| No mention of roaming, which seems to be the biggest hurdle to
| stability and reliability.
|
| Transitioning from one station to another, or worse back to call
| service, seems to consistently causes stability and reliability
| problems.
|
| Worse still, the ability to spread connections across multiple
| spectrums is likely to make the roaming problem worse. 6Ghz will
| usually drop before 5Ghz, and that before 2.4Ghz, which has the
| best wall penetration. How will that impact reliability as people
| walk through buildings?
| eppp wrote:
| We are going to end up with an ap with very low power every
| 400sqft or something the way this is going.
| ipython wrote:
| That's the preferred deployment strategy anyway (less power,
| more cells)
| TaylorAlexander wrote:
| In contrast to the system of wifi access points in
| apartment buildings, which is more cells AND more power!
| konschubert wrote:
| Doesn't work for homes.
| Filligree wrote:
| I don't see why not. That's how I've built mine.
| ikiris wrote:
| Yeah wifi works better when your ap is shouting from
| across your property.
| ipython wrote:
| Absolutely works for homes. Problem is it's more
| expensive and requires know how- it's not just "plug in
| router with max power and 52 antennas"
| TheFuzzball wrote:
| I have a 65sqm apartment with 3 APs, 5GHz set to low power.
| I'm pretty happy with the performance.
| cryptonector wrote:
| > Transitioning from one station to another, or worse back to
| call service
|
| The former is clearly in scope for Wi-Fi protocols. The latter
| is clearly out of scope for Wi-Fi protocols, as it really
| involves work at the IP layer (or higher even).
| Faaak wrote:
| Yeah, this looks to be what Multipath TCP / MPTCP does
| ikiris wrote:
| How is multipath going to tell radios how to roam between
| cells?
| mlyle wrote:
| "Back to cell service" --- you probably have radios
| connected to both wifi and to the cellular network. It's
| not the same radio.
|
| Then you need to decide what links to send frames over
| (and how). MPTCP is one approach to this.
| gsich wrote:
| Roaming is already defined. 802.11k,r,v
| jauntywundrkind wrote:
| 802.11r at least doesn't work on hostapd with wpa3.
| Supposedly it's not compatibleb may threads on that. Although
| there are some counter-examples which I've just found!
| https://forum.openwrt.org/t/802-11r-wpa3-does-it-
| work/135041...
|
| It seems like most clients will stay associated with an AP
| until times are truly truly desperate/signal is near totally
| hosed. As far as I know, the only open source option to
| bandhsteering/ap-steering clients is Dawn, which is quite new
| (relatively speaking). I've been running it on my nighthawk's
| for a while, and it's... Ok. Clients sometimes ping pong
| around, or get sent to a suboptimal band or AP. But it's a
| huge improvement over nothing!
| https://openwrt.org/docs/guide-user/network/wifi/dawn
| throw0101b wrote:
| See the following videos from the Wireless LAN Professionals
| folks:
|
| "IEEE 802 11kvr | Perry Correll | WLPC Phoenix 2019"
|
| * https://www.youtube.com/watch?v=p_K9xHxFM8Y
|
| "Effects of 802.11k/r/v | Andrew McHale | WLPC Prague 2019"
|
| * https://www.youtube.com/watch?v=4Ua2lI6HBhE
|
| "The Challenges of Modern Wi-Fi Roaming Including in 6 GHz |
| Wes Purvis | Prague 2023"
|
| * https://www.youtube.com/watch?v=lAG5jXZCd_8
|
| "Wi-Fi Roaming Revealed"
|
| * https://www.youtube.com/watch?v=4ahOnG76HvQ
|
| The Wifi infrastructure needs to send out the appropriate
| frames (e.g. neighbour reports), and the client has to use
| them to make reasonable decisions.
| PlutoIsAPlanet wrote:
| Wi-Fi 6 was a huge let down, bandwidth increases are useless for
| home users when they require line of sight.
| aendruk wrote:
| The trick I discovered is to have a 500-sqft home. Wi-Fi 6 has
| been great!
| ttyyzz wrote:
| It's true, I still run relatively old Unifi nanoHD APs (3x) and
| considered their newer U6 Pro AP for my house. However, the
| small benefits that would result from this are not worth the
| price to me, especially since everything works great with the
| older gear. We also don't have that many wifi6 clients in the
| family yet, so I discarded the idea.
| karlshea wrote:
| I got a U6 Enterprise the other day, Mac Studio 12 feet away
| with one partial wall in between tries and stays on 6GHz for
| awhile but then falls back to 5GHz pretty quickly.
| gosub100 wrote:
| this, _and_... how many 6 /6e routers _don 't_ try to pressure
| you into "download our app so you can easily control your
| outer" (at the cost of privacy)
| LASR wrote:
| I am always supportive of newer tech. But I am a bit surprised by
| the evolution of standards for both 5G and WiFi over the last few
| years. Compared with the adoption of hardwire ethernet standards
| - which have been seemingly stuck at 1/2.5/10GbE for a while now.
|
| I know there are a ton more benefits of these newer standards,
| but setting those aside for now, are people really feeling
| limited by the speed of 4G LTE & WiFi 5?
|
| I have 6 APs in my home running WiFi 5(AC). My devices usually
| connect between ~800-1300 mbps.
|
| On a few ocassions where I am transferring huge files to / from
| my NAS, I reach for a ethernet adapter.
|
| It's going to cost a lot to upgrade my network, so I've been
| debating whether it's actually worth it or just marketing hype.
| imp0cat wrote:
| The current state of Ethernet in 2023 in consumer devices is
| quite sad. I mean, most TVs and players (and usually not the
| cheap ones, yes, I'm looking at you, Sony) come with 10/100
| Mbit ethernet adapters. I'd expect gigabit to be standard
| everywhere by now.
| barbazoo wrote:
| Where the data taye requires it. No point in putting a
| gigabit adapter in a network speaker. TV, maybe, for 4K and
| beyond.
| gosub100 wrote:
| I can see a point in no longer producing 10mb wired chips
| anymore, just to take the burden off of production. I
| cannot picture a situation where you cannot upgrade to
| 100mb, maybe at the cost of a 48-port 100mb switch, which
| is essentially free these days. at some point it (must)
| cost more to produce antiquated technology, even if your
| equipment (say, a peanut butter factory) doesn't need GigE
| or 100mb bandwidth..
| xxpor wrote:
| Please explain why there needs to be an upgrade.
|
| There's no business case. Even a full blu-ray 4k rip is only
| ~60ish Mbit/sec encoded as HEVC. Netflix and co have been
| pushing streaming bitrates down to sub 10 mbit.
| toast0 wrote:
| 4k uhd blu-ray peak bandwidth is somewhere around 150 mbps.
| Discs can be encoded in different codecs, but h.265 is
| common.
|
| Maybe you can buffer it, but a lot of players lack the
| memory to buffer enough to average out a longer scene with
| intense encoding (such as the parade with confetti in
| Indiana Jones and the Dial of Destiny).
|
| Maybe you can reencode, and AV1 might help a bit, but
| you're likely going to drop some detail in order to fit in
| 100 mbps ethernet.
| pseudosavant wrote:
| The only reason a disc is authored at that bitrate is to
| make bit perfect Blu-ray rips impracticably large. There
| is no visual quality reason to be that high. There is a
| reason Netflix's top quality is a fraction of that
| bitrate and looks excellent.
| ikiris wrote:
| The 3 giant blobs of black on the screen disagree
| strongly.
| pseudosavant wrote:
| The truth is, you need a fairly powerful client (PC, game
| console, flagship smart phone) to ever (occasionally) need
| more than 100mbps Ethernet. There just aren't use cases for
| that much bandwidth.
|
| HD YouTube is still in the extremely low single digit mbps.
| Same for TikTok. Netflix 4K HDR with Atmos easily fits within
| 100mbps. Even 8k streaming will be under 100mbps.
| faeriechangling wrote:
| These boxes often couldn't decode 200mbit of content if they
| tried, and realistically AV1 support is more important than
| if it supports gigabit or not. It's just good economics to
| use a part that already works fine over one that's
| overspecced for the use case.
|
| What are you getting out of it? Faster app and update
| downloads? Those are frills.
| mrandish wrote:
| I bought the recently released top of the line Fire TV 4k Max
| (2nd gen) streaming stick on black friday sale and was
| surprised to discover the external USB port is USB 2.0 speed
| micro-USB. My fault I guess for not even looking closely
| before ordering but I assume these days any brand new, top of
| the line device with a tiny USB port will be USB-C.
|
| Anyway, after disabling the Amazon bloatware and side-loading
| vanilla android apps, even with a fast USB3-capable ethernet
| adapter connected to a 1Gbps RJ-45, standard streaming apps
| wouldn't even stream 4k because the ethernet was too slow. I
| tested the adapter and connection on my laptop and got
| >200mbps which is more than sufficient.
|
| So I was forced to use 5Ghz wifi for the stick to stream 4k,
| thus further clogging up and adding to wifi congestion. BTW,
| I found the same thing to be true with my other streaming
| stick, a Google TV, despite it having a USB-C port. It could
| do 4k streams through an ethernet dongle but it was kind of
| on the edge enough to not always do 4k dependably. But that
| device was >18mo old.
|
| Anyway, it's super annoying that these newish devices are
| making wifi congestion _worse_ by not properly enabling
| reasonable wired streaming (yes, I know there are lots of no-
| name Android TV streaming boxes from Shenzen which have USB-C
| ports with at least USB3.0 support but this new Fire TV stick
| actually has a pretty performant CPU and is otherwise well-
| implemented).
| doublepg23 wrote:
| If you're trying to stream UHD Blu-ray remuxes, you'll have
| best luck with a Shield TV.
| JoshTriplett wrote:
| I love the idea of Wifi 7 and I'm looking forward to upgrading
| from 5/6/6E, for a couple of reasons:
|
| - Yes, I actually do feel limited by the speed of 5 and 6 (not
| 6E), not least of which because it gets worse with distance.
| The 6GHz network works _great_ but doesn 't get nearly enough
| range, and I'd love to supplement that by _simultaneously_
| talking on 5GHz and 2.4GHz, while using as much 6GHz bandwidth
| as is getting through from a given location. And I don 't want
| to have to change which network I'm connected to, or even if
| it's the same ESSID I don't want my device to have to "switch"
| networks.
|
| - No more dealing with 2.4GHz vs 5GHz vs 6GHz channels, or
| worrying which channels are clear, or which devices should be
| on which frequency. The premise of WiFi 7 seems to be to use
| all the bandwidth that nothing else is using, while avoiding
| anything that another device is interfering with. "What channel
| is your router using?" "Yes."
| faeriechangling wrote:
| I feel limited by the latency AND bandwidth of wifi.
|
| Latency interferes with use cases like wireless video streaming
| which is nice for things like Steam Decks or VR. We're not even
| close to being able to decode even gigabit bitrates using
| hardware decoders though, so bandwidth isn't important here. A
| nice coincidence is that Wifi7 coincided with the release of
| hardware av1 encoders/decoders, so I think we're going to see
| these SOCs that support both, and I think that's going to
| marginally improve the viability of computing remotely from the
| edge (Cloud gaming) or home servers (steam link) and cause
| those to be used more heavily.
|
| Bandwidth interferes with large sequential file transfers, and
| WANs are starting to have more bandwidth than wifi, so this is
| starting to cramp things like large downloads of movie content
| or .iso's or cloud backups. It also bottlenecks 2.5gbe LANs
| which are economical given they run on cat5e networks.
|
| I think wifi7 is going to explode as ISPs bundle wifi7 modems
| with their higher bandwidth plans, so it's going to be pushed
| because the ISPs need it more than because most customers need
| it. I only expect certain pros, gamers, and large scale
| commercial installers to invest in wifi7 themselves.
| klabb3 wrote:
| To add to that: throughput isn't just about bandwidth, but
| also latency (BDP) but more importantly (in the case of WiFi)
| jitter and packet loss.
|
| Even if your WiFi supports 1Gbit most of the time, tcp will
| throttle back to slow start (IIRC - the linear one) upon
| packet loss. And time to recovery depends on RTT. When I
| measure I always get inconsistent results on WiFi, whereas
| ethernet+fiber is butter smooth.
| afavour wrote:
| My understanding was that 5G isn't just a speed improvement,
| it's latency etc as well. But, no, I don't feel limited by my
| cellphone being on LTE. That being said 5G opens a lot of
| doors: my home internet connection is 5G now, I pull around
| 500Mbps with not particularly awful/noticeable latency.
| coolspot wrote:
| > I have 6 APs in my home running WiFi 5(AC).
|
| How does that work? Are they components of a mesh network
| bundle? Do they have same SSID? Are they aware of each other?
| LASR wrote:
| They are all Unifi APs from 2 different upgrade cycles. 3x
| UnFi AC Pro and 3x Unifi NanoHD.
|
| After I got the newer NanoHDs, I just kept the older AC Pros
| around for 2.4Ghz devices and exterior coverage.
|
| Same SSID and they are all hooked up the same controller etc,
| so clients can roam.
| pseudosavant wrote:
| I have three 5/AC access points in my house. All broadcast
| the same two SSIDs: 2.4Ghz+5Ghz and 5Ghz only). I have
| 802.11r Fast Transition enabled in OpenWRT to improve
| roaming, but have never quantified a benefit. Each AP is set
| on different channels to minimize overlap.
| davkan wrote:
| In the enterprise it's a combination of 802.11k 802.11v and
| 802.11r.[0] All APs have the same ssid, clients initiate
| transferring between APs in the background based on signal
| strength, silently re-authenticating to the new ap.
|
| Many (most?) off-the-shelf home wifi mesh systems dont
| implement 802.11r so you may have a brief interruption during
| transfer, though it still happens without user input. Some
| older/embedded devices may have issues connecting with
| 802.11r enabled networks as well.
|
| Off-the-shelf home mesh systems typically have one mesh node
| with all the functions and webui of a normal SoHo router, and
| then other nodes are in AP mode only with either wired or
| wireless backhaul to the main node. Enterprise setups instead
| will have a separate controller for the APs. Also typically
| all APs in an enterprise will be bridged to the same L2
| networks whereas some home mesh systems will use a separate
| broadcast domain for each AP.
|
| [0] https://learn.microsoft.com/en-us/windows-
| hardware/drivers/n...
| orev wrote:
| WiFi is used in places other than homes, and many parts of the
| newer standards are focused on those situations. Hotels,
| trains, conference centers, stadiums, offices, apartment
| buildings, etc. all need to be able to handle large numbers of
| users. If what you have is working for your case, there's no
| reason to upgrade.
| toast0 wrote:
| > are people really feeling limited by the speed of 4G LTE &
| WiFi 5?
|
| 4G LTE IMHO, is fine for mobile use, but wouldn't really be
| enough for a household fixed wireless connection; 5G is much
| better for that. Occasionaly I run into what seems to be
| congested towers, and denser encoding should help with that
| (unless the problem is the backhaul). Of course, additional
| spectrum allocations is likely the most useful part of 5g,
| although that didn't have to be tied.
|
| Also IMHO, WiFi 5 is limited by running only on 5 GHz, which
| doesn't go far enough. I'm holding out for now, but I'm fairly
| sure I'll have significiantly better experiences with wifi 6
| than wifi 5/4. Being able to run the newest encoding on 2.4Ghz
| will be great for me... but I live in the middle of the woods
| and don't see a lot of outside APs. I'm hoping to find APs that
| are cheap and let me enhance roaming.
| amluto wrote:
| > Imagine a wireless factory robot in a situation where a worker
| suddenly steps in front of it and the robot needs to make an
| immediate decision.
|
| If there is a factory that allows a worker to step in front of a
| robot and depends on a wireless communication mechanism for
| safety, something is very very wrong.
| TaylorAlexander wrote:
| Yeah that is a horrible example.
| jandrese wrote:
| I guess it has to call back to the manufacturer's cloud service
| to see if the factory subscribed to the "avoid crushing people"
| option.
| jauntywundrkind wrote:
| Discarding ideas out of hand is easy & cheap. Why do you think
| this so absolutely unacceptable an idea?
|
| A factory seems like a big space where there could perhaps be a
| predictable well known radio spectrum, and where that spectrum
| could reliably be monitored for stability. Unlike most spaces,
| there's a lot of possibilities for controlling the environment
| well, if you wanted; no end-user devices allowed, nothing but
| known systems.
|
| This feels like you could very quickly get to some kind of
| known state. "We will be able to send priority signals within
| under 2ms." Or some such quality metric. Keep a signal analyzer
| going: any time this isn't met, everything goes into an
| independent failsafe shutdown mode until the issue is debugged
| and started up again.
|
| Our instincts to judge & discard ideas & possibilities is _so_
| dangerous, is such a deep human peril. Your instinct feels like
| it comes from a good place, feels correct, but ultimately, this
| is acting in fear uncertainty & doubt, & your execution rests
| on no contentions other than it's own self certainty.
|
| I think we must be much more thorough in our thinking &
| arguing, must not leap to discard ideas so quickly totally &
| casually; FUD must be rooted out deep within ourselves &
| challenged. Let not progress be onstructed by meek wills.
| 0cf8612b2e1e wrote:
| I think it is easy to dismiss because you should always keep
| squishy humans away from the unthinking machines. If they are
| always separated during routine operation, you have
| eliminated potential failure modes that could result in a
| fatal injury.
|
| Sure, engineer the machines to immediately halt if a human is
| detected, but also design the system so that close physical
| proximity is never required.
| MR4D wrote:
| Have you ever been to a modern warehouse?
|
| There are hundreds of those things moving around, often with
| heavy loads (enough to injure if not enough to kill).
|
| And it has been that way for years. [0]
|
| [0] - https://www.wired.com/story/amazon-warehouse-robots/
|
| EDIT : Adding Amazon's page on robots that also talks about
| safety: https://www.amazon.science/latest-news/amazon-robotics-
| auton...
| Unfrozen0688 wrote:
| lol ethernet to everything please. soooo may helldesk calls.
| "uoooh internet... bad..." okay plug dat shiiiii in
| ttyyzz wrote:
| What? Of course you should wire everything up if possible, but
| some devices just don't allow that. I once used the USB to
| Ethernet adapter that came with my Fire TV, only to discover
| that it runs at 100 Mbit and that the USB controller is a
| bottleneck. In addition, the connection kept dropping out, so I
| switched back to WiFi for that specific device.
| im3w1l wrote:
| This sounds like a good direction: even very infrequent, very
| brief stutters can be a dealbreaker for hardcore gamers, and make
| them chose wired over wifi.
| cryptonector wrote:
| > "The question that we posed ourselves was, 'What do we do
| now?'" says Carlos Cordeiro, [...]
|
| Uh, Wi-Fi security still sucks. Can we fix that?
| mschuster91 wrote:
| Does it?
|
| My only gripe is that for whatever goddamn reason Android isn't
| able to detect what specific subset of WPA2 Enterprise config
| the AP actually wants/supports and the UI seems to constantly
| change around over the years, a common annoyance in large
| campus networks where the admins don't stay on top.
| transpute wrote:
| Wi-Fi 7 access point firmware vendors could choose to implement
| optional per-client passphrases, VLANs and device identity for
| access and traffic control. OSS reference code is available
| from https://github.com/spr-networks/super.
|
| Wi-Fi 7 Sensing (802.11bf) of humans can measure breathing
| rate, keystroke typing, position and motion, from outside the
| walls of homes and business. What's the relationship between
| human biometric recognition and security?
| bofaGuy wrote:
| I cannot remember the last time WiFi reliably was an issue for
| me. Maybe over 5 years ago..?
| jl6 wrote:
| Industry, please please please take your time and get the feature
| set right and stick to calling it simply WiFi 7 - or just leave
| the marginal stuff for WiFi 8. I don't want to see a 7E.
___________________________________________________________________
(page generated 2023-12-29 23:01 UTC)