From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-yw1-x112e.google.com (mail-yw1-x112e.google.com [IPv6:2607:f8b0:4864:20::112e]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by lists.bufferbloat.net (Postfix) with ESMTPS id 2B81A3CBC5; Wed, 15 Mar 2023 18:08:43 -0400 (EDT) Received: by mail-yw1-x112e.google.com with SMTP id 00721157ae682-5418c2b8ef2so227920727b3.5; Wed, 15 Mar 2023 15:08:43 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112; t=1678918122; h=cc:to:subject:message-id:date:from:in-reply-to:references :mime-version:from:to:cc:subject:date:message-id:reply-to; bh=fsJ8teHX6+hIovgWcuoffuj/Tw/fICjrlInLlxqJ47s=; b=CEJohCzKCW2MHiV6evWNt2ww62b3An+oZkjwUSfKkIlYqZfjkYO6inFubnxsJwvIt5 uZ8oi17YfR+1wqXgsgzcdAoRpc9gPKBjWul7X5R0Pq5rcnLumHse1JEzK3znHA3QIw/j /cG+QcavOFFi0VnJTkUAPjp2cWxkRELQNKrspvVEcfqizId0OziIlPr4vq2W2miPLMx9 kS/49X6kFOTkBVH9Cj2CG+fbIQ41zSyuvPnIy0DyWTovXEVvbL9dHJJEPWFGFvLTs18T X7ebnmZ9J+7u7vQ/Rg+6qnb7Yk46a/+SUdm42R8pfvX6gKFRMWo2F0/OrEV/7arWOVzs Ki5Q== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; t=1678918122; h=cc:to:subject:message-id:date:from:in-reply-to:references :mime-version:x-gm-message-state:from:to:cc:subject:date:message-id :reply-to; bh=fsJ8teHX6+hIovgWcuoffuj/Tw/fICjrlInLlxqJ47s=; b=0AN6pePulr1MC3QHcE8flsbdsogCTnqoEJ6l5tFvN9+lS6rQMF8JY9yR5JLZG8Xf9t nB7tn9mooj7a2enqa9YQoMjpHcivjjXPGMOraNoUffpR3XlSDY23qoBxob2EDVD++fP3 yJb9jE8rNCwS9kH8gCDM/ODq+4d+bWB0sbBfbtsefRKbLCA+da0W6oVQH9EQT+qTzviA xEsD5fgSt3li0aPd8HLz0KHZ7DJSVMiqEJZK1BduFgdoUexcguqE7RuPY1zgz9LXcraF AIqylVQ9TldITo3iY408OQBnCiivSSiokIRe6aGDcQsbNfltegNRrCB/7yzpFXhfVh6F +0Vg== X-Gm-Message-State: AO0yUKUaJhJ3McylQgDNpUGlNdi8pNEXRlqMK/fDV7JiGmuf/rlyPN05 0HPb5FKsRtgxHNzcPXQQVNPsljSWCGdvoHhzICbx8yNx X-Google-Smtp-Source: AK7set8HyOAaQbO22RH3lRkvy7HX4JYun7rdrHJ8v9tP9JvEJ7OIRZ9TEEnZb1BccthPSW+AgNA3kKw37kkjkH0XAV4= X-Received: by 2002:a81:4517:0:b0:536:38b4:f50 with SMTP id s23-20020a814517000000b0053638b40f50mr942914ywa.1.1678918122397; Wed, 15 Mar 2023 15:08:42 -0700 (PDT) Received: from 1064022179695 named unknown by gmailapi.google.com with HTTPREST; Wed, 15 Mar 2023 18:08:41 -0400 Received: from 1064022179695 named unknown by gmailapi.google.com with HTTPREST; Wed, 15 Mar 2023 18:08:38 -0400 Mime-Version: 1.0 (Mimestream 0.41.5) References: <89c55d67-86f0-494d-a09e-c9aeebe46dc0@rjmcmahon.com> <70CBB03C-4394-4A93-BBB5-7449DC1AAF9C@gmx.de> <063359bf-5bf3-4688-852c-a7d81e6b80a3@rjmcmahon.com> <21f2252ff57e60dc52e7b9a6db8ba936@rjmcmahon.com> <8a04de1c-2d47-4226-a1ac-ea3d5e7b7253@rjmcmahon.com> <949886c0f65025a626887d09a818c233@rjmcmahon.com> <48q03o08-ro20-2p59-1644-sp73526024q3@ynat.uz> In-Reply-To: From: dan Date: Wed, 15 Mar 2023 18:08:41 -0400 Message-ID: To: Dave Taht Cc: Rpm , libreqos , Bruce Perens , Dave Taht via Starlink , bloat , David Lang Content-Type: multipart/alternative; boundary="0000000000004ae7e105f6f796a1" Subject: Re: [Starlink] [Rpm] [Bloat] [LibreQoS] On FiWi X-BeenThere: starlink@lists.bufferbloat.net X-Mailman-Version: 2.1.20 Precedence: list List-Id: "Starlink has bufferbloat. Bad." List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 15 Mar 2023 22:08:43 -0000 --0000000000004ae7e105f6f796a1 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable On Mar 15, 2023 at 4:04:27 PM, Dave Taht wrote: > On Wed, Mar 15, 2023 at 2:52=E2=80=AFPM David Lang wrote: > > > On Wed, 15 Mar 2023, Dave Taht wrote: > > > > On Wed, Mar 15, 2023 at 12:33=E2=80=AFPM David Lang via Rpm > > > wrote: > > >> > > >> if you want another example of the failure, look at any conference > center, they > > >> have a small number of APs with wide coverage. It works well when the > place is > > >> empty and they walk around and test it, but when it fills up with > users, the > > >> entire network collapses. > > >> > > >> Part of this is that wifi was really designed for sparse environments, > so it's > > >> solution to "I didn't get my message through" is to talk slower (and > louder if > > >> possible), which just creates more interference for other users and > reduces the > > >> available airtime. > > >> > > >> I just finished the Scale conference in Pasadena, CA. We deployed over > 100 APs > > >> for the conference, up to 7 in a room, on the floor (so that the > attendees > > >> bodies attenuate the signal) at low power so that the channels could b= e > re-used > > >> more readily. > > > > > > How did it go? You were deploying fq_codel on the wndr3800s there as > > > of a few years ago, and I remember you got rave reviews... (can you > > > repost the link to that old data/blog/podcast?) > > > no good stats this year. still using the wndr3800s. Lots of people > commenting on > > how well the network did, but we were a bit behind this year and didn't > get good > > monitoring in place. No cake yet. > > > I think this is what you mean > > https://www.youtube.com/watch?v=3DUXvGbEYeWp0 > > > > A point I would like to make for the africa contingent here is that > you do not need the latest > technology for africa. We get 300Mbit out of hardware built in the > late 00s, like the wndr3800. The ath9k chipset is STILL manufactured, > the software mature, and for all I know millions of routers > like these are lying in junk bins worldwide, ready to be recycled and > reflashed. > > One libreqos customer deployed libreqos, and took a look at the 600+ > ubnt AGWs (ath9k based), on the shelf that could be fq_codeled, > especially on the wifi... built a custom openwrt imagebuilder image > for em, reflashed and redistributed them. > > The wndr3800s were especially well built. I would expect them to last > decades. I had one failure of one that had been in the field for over > 10 years... I thought it was the flash chip... no, it was the power > supply! > > > > Did you get any good stats? > > > > > > Run cake anywhere? > > >> > > >> in the cell phone world they discovered 'microcells' years ago, but > with wifi > > >> too many people are still trying to cover the max area with the fewest > possible > > >> number of radios. As Dan says, it just doesn't work. > > >> > > >> and on mesh radios, you need to not just use a different channel for > your > > >> uplink, you need a different band to avoid desense on the connection t= o > your > > >> users. And that uplink is going to have the same hidden transmitter an= d > airtime > > >> problems competing with the other nodes also doing the uplink that it'= s > > >> scalability is very limited (even with directional antennas). > Wire/fiber for the > > >> uplink is much better. > > >> > > >> David Lang > > >> > > >> > > >> > > >> On Wed, 15 Mar > > >> 2023, dan via Bloat wrote: > > >> > > >>> Trying to do all of what is currently wanted with 1 AP in a house is = a > huge > > >>> part of the current problems with WiFi networks. MOAR power to try t= o > > >>> overcome attenuation and reflections from walls so more power bleeds > into > > >>> the next home/suite/apartment etc. > > >>> > > >>> In the MSP space it's been rapidly moving to an AP per room with outp= ut > > >>> turned down to minimum. Doing this we can reused 5Ghz channels 50f= t > away > > >>> (through 2 walls etc...) without interference. > > >>> > > >>> One issue with the RRH model is that to accomplish this 'light bulb' > model, > > >>> ie you put a light bulb in the room you want light, is that it requir= es > > >>> infrastructure cabling. 1 RRH AP in a house is already a failure > today and > > >>> accounts for most access complaints. > > >>> > > >>> Mesh radios have provided a bit of a gap fill, getting the access SSI= D > > >>> closer to the device and backhauling on a separate channel with bette= r > (and > > >>> likely fixed position ) antennas. > > >>> > > >>> regardless of my opinion on the full on failure of moving firewall of= f > prem > > >>> and the associated security risks and liabilities, single AP in a hom= e > is > > >>> already a proven failure that has given rise to the mesh systems that > are > > >>> top sellers and top performers today. > > >>> > > >>> IMO, there was a scheme that gained a moment of fame and then died ou= t > of > > >>> powerline networking and an AP per room off that powerline network. = I > have > > >>> some of these deployed with mikrotik PLA adapters and the model works > > >>> fantastically, but the powerline networking has evolved slowly so I'm > > >>> seeing ~200Mbps practical speeds, and the mikrotik units have 802.11n > > >>> radios in them so also a bit of a struggle for modern speeds. This > model, > > >>> with some development to get ~2.5Gbps practical speeds, and WiFi6 or > WiFi7 > > >>> per room at very low output power, is a very practical and deployable > by > > >>> consumers setup. > > >>> > > >>> WiFi7 also solves some pieces of this with AP coordination and > > >>> co-transmission, sort of like a MUMIMO with multiple APs, and that's = in > > >>> early devices already (TPLINK just launched an AP). > > >>> > > >>> IMO, too many hurdles for RRH models from massive amounts of > unfrastructure > > >>> to build, homes and appartment buildings that need re-wired, security > and > > >>> liability concerns of homes and business not being firewall isolated = by > > >>> stakeholders of those networks. > > >>> > > >>> On Wed, Mar 15, 2023 at 11:32=E2=80=AFAM rjmcmahon > wrote: > > >>> > > >>>> The 6G is a contiguous 1200MhZ. It has low power indoor (LPI) and ve= ry > > >>>> low power (VLP) modes. The pluggable transceiver could be color code= d > to > > >>>> a chanspec, then the four color map problem can be used by installer= s > > >>>> per those chanspecs. https://en.wikipedia.org/wiki/Four_color_theore= m > > >>>> > > >>>> There is no CTS with microwave "interference" The high-speed PHY rat= es > > >>>> combined with low-density AP/STA ratios, ideally 1/1, decrease the > > >>>> probability of time signal superpositions. The goal with wireless > isn't > > >>>> high densities but to unleash humans. A bunch of humans stuck in a d= og > > >>>> park isn't really being unleashed. It's the ability to move from blo= ck > > >>>> to block so-to-speak. FiWi is cheaper than sidewalks, sanitation > > >>>> systems, etc. > > >>>> > > >>>> The goal now is very low latency. Higher phy rates can achieve that > and > > >>>> leave the medium free the vast most of the time and shut down the RR= H > > >>>> too. Engineering extra capacity by orders of magnitude is better tha= n > > >>>> AQM. This has been the case in data centers for decades. Congestion? > Add > > >>>> a zero (or multiple by 10) > > >>>> > > >>>> Note: None of this is done. This is a 5-10 year project with zero > > >>>> engineering resources assigned. > > >>>> > > >>>> Bob > > >>>>> On Tue, Mar 14, 2023 at 5:11=E2=80=AFPM Robert McMahon > > >>>>> wrote: > > >>>>> > > >>>>>> the AP needs to blast a CTS so every other possible conversation h= as > > >>>>>> to halt. > > >>>>> > > >>>>> The wireless network is not a bus. This still ignores the hidden > > >>>>> transmitter problem because there is a similar network in the next > > >>>>> room. > > >>>> > > >>> _______________________________________________ > > >> Bloat mailing list > > >> Bloat@lists.bufferbloat.net > > >> https://lists.bufferbloat.net/listinfo/bloat > > >> _______________________________________________ > > >> Rpm mailing list > > >> Rpm@lists.bufferbloat.net > > >> https://lists.bufferbloat.net/listinfo/rpm > > > > > > > > > > > > > > > > > -- > Come Heckle Mar 6-9 at: https://www.understandinglatency.com/ > Dave T=C3=A4ht CEO, TekLibre, LLC > Much of the hardware dumped on the US market in particular is especially poorly made. Ie, engineered for our disposable market. Lots of netgear products for example have a typical usable life of just 2-3 years if that, and then the caps have busted or some patina on the boards has killed them. I know Europe has some standards on this as well as South Korea to give them longer life. To the point, it=E2=80=99s not realistic to recycle thes= e items from the US to other place because they were =E2=80=98built to fail=E2=80= =99. --0000000000004ae7e105f6f796a1 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable


On Mar 15, 2023 at 4:04:27 PM, Da= ve Taht <dave.taht@gmail.com&= gt; wrote:
=20
On Wed, Mar 15, 2023 at 2:52=E2=80=AFPM David Lang <david@lang.hm> wrote:
=
On Wed, 15 Mar 2023, Dave Taht = wrote:

> On Wed, Mar 15, 2023 at 12:33=E2=80=AFPM David Lang = via Rpm
> <rpm@lists.bufferbloat.net> wrote:
<= /blockquote>
>>
>> if you want another example of the failure, look a= t any conference center, they
&g= t;> have a small number of APs with wide coverage. It works well when th= e place is
>> empty and th= ey walk around and test it, but when it fills up with users, the
>> entire network collapses.
>>
>> Part of this is that wifi was really designed for spa= rse environments, so it's
&g= t;> solution to "I didn't get my message through" is to ta= lk slower (and louder if
>>= ; possible), which just creates more interference for other users and reduc= es the
>> available airtim= e.
>>
>> I just finished the Scale conference in Pas= adena, CA. We deployed over 100 APs
>> for the conference, up to 7 in a room, on the floor (so that t= he attendees
>> bodies att= enuate the signal) at low power so that the channels could be re-used
>> more readily.
>
> How did it go? You were deploying fq_codel on the wndr3800s there as=
> of a few years ago, and I = remember you got rave reviews... (can you
> repost the link to that old data/blog/podcast?)

= no good stats this year. still using the wndr3800s. Lots of people commenti= ng on
how well the network did, = but we were a bit behind this year and didn't get good
=
monitoring in place. No cake yet.

I t= hink this is what you mean
https://www.youtube.co= m/watch?v=3DUXvGbEYeWp0


A point I would like to= make for the africa contingent here is that
you do not need the latest<= br>technology for africa. We get 300Mbit out of hardware built in the
la= te 00s, like the wndr3800. The ath9k chipset is STILL manufactured,
the = software mature, and for all I know millions of routers
like these are l= ying in junk bins worldwide, ready to be recycled and
reflashed.

= One libreqos customer deployed libreqos, and took a look at the 600+
ubn= t AGWs (ath9k based), on the shelf that could be fq_codeled,
especially = on the wifi... built a custom openwrt imagebuilder image
for em, reflash= ed and redistributed them.

The wndr3800s were especially well built.= I would expect them to last
decades. I had one failure of one that had = been in the field for over
10 years... I thought it was the flash chip..= . no, it was the power
supply!


>= ; Did you get any good stats?
&g= t;
> Run cake anywhere?
>>
>> in the cell phone world they discovered 'microc= ells' years ago, but with wifi
>> too many people are still trying to cover the max area with the= fewest possible
>> number= of radios. As Dan says, it just doesn't work.
>>
>= > and on mesh radios, you need to not just use a different channel for y= our
>> uplink, you need a = different band to avoid desense on the connection to your
<= blockquote type=3D"cite"> >> users. And that uplink is going to have = the same hidden transmitter and airtime
>> problems competing with the other nodes also doing the upl= ink that it's
>> scala= bility is very limited (even with directional antennas). Wire/fiber for the=
>> uplink is much better.=
>>
>> David Lang
>>
>>
=
>>
>> =C2=A0=C2=A0On Wed, 15 Mar
>> 2023, dan via Bloat wrote:
>>
>>> Trying to do all of what is currently wanted with 1 AP in a = house is a huge
>>> par= t of the current problems with WiFi networks.=C2=A0 MOAR power to try to
>>> overcome attenuation = and reflections from walls so more power bleeds into
>>> the next home/suite/apartment etc.
>>>
>>> In the MSP space it's been rapidly movin= g to an AP per room with output
= >>> turned down to minimum. =C2=A0=C2=A0=C2=A0Doing this we can re= used 5Ghz channels 50ft away
>= ;>> (through 2 walls etc...) without interference.
>>>
>>> One issue with the RRH model is that to accomplish this = 'light bulb' model,
>= >> ie you put a light bulb in the room you want light, is that it req= uires
>>> infrastructur= e cabling. =C2=A01 RRH AP in a house is already a failure today and
>>> accounts for most access c= omplaints.
>>>
>>> Mesh radios have provided a= bit of a gap fill, getting the access SSID
>>> closer to the device and backhauling on a separate= channel with better (and
>&g= t;> likely fixed position ) antennas.
>>>
>>= > regardless of my opinion on the full on failure of moving firewall off= prem
>>> and the assoc= iated security risks and liabilities, single AP in a home is
>>> already a proven failure that has= given rise to the mesh systems that are
>>> top sellers and top performers today.
>>>
>>> IMO, there was a scheme that gained a moment of fam= e and then died out of
>>&= gt; powerline networking and an AP per room off that powerline network.=C2= =A0 I have
>>> some of = these deployed with mikrotik PLA adapters and the model works
>>> fantastically, but the powerline= networking has evolved slowly so I'm
>>> seeing ~200Mbps practical speeds, and the mikrotik = units have 802.11n
>>> = radios in them so also a bit of a struggle for modern speeds. =C2=A0=C2=A0T= his model,
>>> with som= e development to get ~2.5Gbps practical speeds, and WiFi6 or WiFi7
>>> per room at very low output= power, is a very practical and deployable by
>>> consumers setup.
>>>
>>= ;> WiFi7 also solves some pieces of this with AP coordination and
>>> co-transmission, sort of = like a MUMIMO with multiple APs, and that's in
>>> early devices already (TPLINK just launched= an AP).
>>>
>>> IMO, too many hurdles for RRH= models from massive amounts of unfrastructure
>>> to build, homes and appartment buildings that n= eed re-wired, security and
>&= gt;> liability concerns of homes and business not being firewall isolate= d by
>>> stakeholders o= f those networks.
>>>
>>> On Wed, Mar 15, 2023= at 11:32=E2=80=AFAM rjmcmahon <rjmcmahon@rjmcmahon.com> wrote:
>>>
>>= >> The 6G is a contiguous 1200MhZ. It has low power indoor (LPI) and = very
>>>> low power = (VLP) modes. The pluggable transceiver could be color coded to
>>>> a chanspec, then the four c= olor map problem can be used by installers
>>>> per those chanspecs. https://en.wikipedia.org/wiki/Four_color= _theorem
>>>>
>>>> There is no CTS w= ith microwave "interference" The high-speed PHY rates
>>>> combined with low-density = AP/STA ratios, ideally 1/1, decrease the
>>>> probability of time signal superpositions. The = goal with wireless isn't
>= ;>>> high densities but to unleash humans. A bunch of humans stuck= in a dog
>>>> park = isn't really being unleashed. It's the ability to move from block
>>>> to block so-to-s= peak. FiWi is cheaper than sidewalks, sanitation
>>>> systems, etc.
>>>>
= >>>> The goal now is very low latency. Higher phy rates can ac= hieve that and
>>>> = leave the medium free the vast most of the time and shut down the RRH
>>>> too. Engineering ext= ra capacity by orders of magnitude is better than
>>>> AQM. This has been the case in data cent= ers for decades. Congestion? Add
= >>>> a zero (or multiple by 10)
>>>>
&g= t;>>> Note: None of this is done. This is a 5-10 year project with= zero
>>>> engineeri= ng resources assigned.
>>&= gt;>
>>>> Bob
=
>>>>> On Tue, Mar 14= , 2023 at 5:11=E2=80=AFPM Robert McMahon
>>>>> <rjmcmahon@rjmcmahon.com> wrote:
>>>>>
= >>>>>> the AP needs to blast a CTS so every other possibl= e conversation has
>>>&= gt;>> to halt.
>>>= ;>>
>>>>> T= he wireless network is not a bus. This still ignores the hidden
>>>>> transmitter problem be= cause there is a similar network in the next
>>>>> room.
>>>>
>>= ;> _______________________________________________
>> Bloat mailing list
>> B= loat@lists.bufferbloat.net
&= gt;> https://li= sts.bufferbloat.net/listinfo/bloat
>> _______________________________________________
>> Rpm mailing list
<= blockquote type=3D"cite"> >> Rpm@lists.bufferbloat.net
>> https://= lists.bufferbloat.net/listinfo/rpm
>
>
>
= >



--
Come Heckle Mar 6-9 at: https://www.understandinglatency.co= m/
Dave T=C3=A4ht CEO, TekLibre, LLC

Much of the hardware dump= ed on the US market in particular is especially poorly made.=C2=A0 Ie, engi= neered for our disposable market.=C2=A0 Lots of netgear products for exampl= e have a typical usable life of just 2-3 years if that, and then the caps h= ave busted or some patina on the boards has killed them. =C2=A0

I know Europe has some standards on t= his as well as South Korea to give them longer life.=C2=A0 To the point, it= =E2=80=99s not realistic to recycle these items from the US to other place = because they were =E2=80=98built to fail=E2=80=99.

--0000000000004ae7e105f6f796a1--