Re: [Cerowrt-devel] Fwd: wndr3800 replacement

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Martin Bailey]

The recently released v2 of the TP-Link Archer C7 router would be a
pretty good low-cost replacement. It's fully open-source Atheros/Qualcomm,
includes a fairly high performance 720MHz SOC with 128MB RAM, 16MB
flash and 6 dual-band antennas (AC1750) with very good range. It can
be found for $99 right now. The first hardware revision isn't
supported by the ath10k driver in OpenWRT so make sure to only
consider v2.

http://wiki.openwrt.org/toh/tp-link/tl-wdr7500
https://wikidevi.com/wiki/TP-LINK_Archer_C7_v2.x
http://www.newegg.com/Product/Product.aspx?Item=N82E16833704177

> On Thu, 27 Mar 2014, Aaron Wood wrote:
>
>> Date: Thu, 27 Mar 2014 13:50:27 +0100
>> From: Aaron Wood <woody77 at gmail.com>
>> To: David Lang <david at lang.hm>
>> Cc: Dave Taht <dave.taht at gmail.com>,
>>     "cerowrt-devel at lists.bufferbloat.net"
>>     <cerowrt-devel at lists.bufferbloat.net>
>> Subject: Re: [Cerowrt-devel] Fwd: wndr3800 replacement
>>
>> On Wed, Mar 26, 2014 at 11:11 PM, David Lang <david at lang.hm> wrote:
>>
>>> If the openwrt folks could figure out how they are going to deal with
>>> NAND
>>> flash, it would be nice to be able to use one of the many routers that is
>>> shipping with more flash (128M in the newer netgear routers would be
>>> nice)
>>>
>>> if I were to get my hands on one, what sort of testing would you want to
>>> do to it to tell if it looks like it would hold up?
>>
>>
>> I have experience running mtd on NAND, using jffs2.  It seems to be
>> holding
>> up well.  Better than NOR did, honestly.  Although in general, I wish they
>> would shift to eMMC.  But it's driven by two factors:
>>
>> 1) part cost
>> 2) chipset support from the router SoC vendors
>>
>> Given some of the wishes that I see on here, I think for development,
>> people would be happier with a platform that wasn't based on a router SoC
>> (like the wndr is), but instead was based on an embedded application
>> processor with PCIe for the radios, and an external switch fabric.
>
> I think we have two competing desires.
>
> one is to have a nice powerful device for those people who have fast
> connections
> and for us to experiment with.
>
> the second is to have a 'home' device.
>
> using a 3800 or similarly priced ($100-$150 USD) device that's readily
> available
> is very good for the second category, the question is if we can find one
> that's
> powerful enough for the first.
>
> David Lang
>
>>  But for
>> thermal purposes alone, I've been seeing more and more external switch
>> fabrics.  The heat of a 5-port gigabit switch IC is pretty substantial
>> (from my teardowns).
>>
>> One item I think will be a boon, especially with DNSSEC, is super-cap or
>> battery-backed rtc, but that's asking for a unicorn, I think.  Or...  a
>> Gateworks Ventana GW5310 loaded with a couple standard (industrial-grade)
>> PCIe radios, loaded into a custom case.  My guess is that it's a pretty
>> expensive route, though.  I would be surprised if a completely assembled
>> unit would be <$300.  At which point it starts to look better to just run
>> a
>> separate router and AP (using standard wndr-type platforms as the APs and
>> a
>> higher-end board or PC as the gateway).
>>
>> -Aaron
>>

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Dave Taht]

On Sat, Mar 29, 2014 at 12:27 PM, Martin Bailey <martin@pcalpha.com> wrote:
> The recently released v2 of the TP-Link Archer C7 router would be a
> pretty good low-cost replacement. It's fully open-source Atheros/Qualcomm,
> includes a fairly high performance 720MHz SOC with 128MB RAM, 16MB
> flash and 6 dual-band antennas (AC1750) with very good range. It can
> be found for $99 right now. The first hardware revision isn't
> supported by the ath10k driver in OpenWRT so make sure to only
> consider v2.

Boy is that a big mini-pci card! (won't fit in most mini-pci slots)

I'm not very happy with the ath10k right now but it HAS been getting better.

Is there BQL support yet for the ethernet chip?

955x_GMAC: qca955x_soc_gmac_set_mac_duplex 955x_GMAC:
qca955x_soc_gmac_set_link Done


>
> http://wiki.openwrt.org/toh/tp-link/tl-wdr7500
> https://wikidevi.com/wiki/TP-LINK_Archer_C7_v2.x
> http://www.newegg.com/Product/Product.aspx?Item=N82E16833704177
>
>> On Thu, 27 Mar 2014, Aaron Wood wrote:
>>
>>> Date: Thu, 27 Mar 2014 13:50:27 +0100
>>> From: Aaron Wood <woody77 at gmail.com>
>>> To: David Lang <david at lang.hm>
>>> Cc: Dave Taht <dave.taht at gmail.com>,
>>>     "cerowrt-devel at lists.bufferbloat.net"
>>>     <cerowrt-devel at lists.bufferbloat.net>
>>> Subject: Re: [Cerowrt-devel] Fwd: wndr3800 replacement
>>>
>>> On Wed, Mar 26, 2014 at 11:11 PM, David Lang <david at lang.hm> wrote:
>>>
>>>> If the openwrt folks could figure out how they are going to deal with
>>>> NAND
>>>> flash, it would be nice to be able to use one of the many routers that is
>>>> shipping with more flash (128M in the newer netgear routers would be
>>>> nice)
>>>>
>>>> if I were to get my hands on one, what sort of testing would you want to
>>>> do to it to tell if it looks like it would hold up?
>>>
>>>
>>> I have experience running mtd on NAND, using jffs2.  It seems to be
>>> holding
>>> up well.  Better than NOR did, honestly.  Although in general, I wish they
>>> would shift to eMMC.  But it's driven by two factors:
>>>
>>> 1) part cost
>>> 2) chipset support from the router SoC vendors
>>>
>>> Given some of the wishes that I see on here, I think for development,
>>> people would be happier with a platform that wasn't based on a router SoC
>>> (like the wndr is), but instead was based on an embedded application
>>> processor with PCIe for the radios, and an external switch fabric.
>>
>> I think we have two competing desires.
>>
>> one is to have a nice powerful device for those people who have fast
>> connections
>> and for us to experiment with.
>>
>> the second is to have a 'home' device.
>>
>> using a 3800 or similarly priced ($100-$150 USD) device that's readily
>> available
>> is very good for the second category, the question is if we can find one
>> that's
>> powerful enough for the first.
>>
>> David Lang
>>
>>>  But for
>>> thermal purposes alone, I've been seeing more and more external switch
>>> fabrics.  The heat of a 5-port gigabit switch IC is pretty substantial
>>> (from my teardowns).
>>>
>>> One item I think will be a boon, especially with DNSSEC, is super-cap or
>>> battery-backed rtc, but that's asking for a unicorn, I think.  Or...  a
>>> Gateworks Ventana GW5310 loaded with a couple standard (industrial-grade)
>>> PCIe radios, loaded into a custom case.  My guess is that it's a pretty
>>> expensive route, though.  I would be surprised if a completely assembled
>>> unit would be <$300.  At which point it starts to look better to just run
>>> a
>>> separate router and AP (using standard wndr-type platforms as the APs and
>>> a
>>> higher-end board or PC as the gateway).
>>>
>>> -Aaron
>>>
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel



-- 
Dave Täht

Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Dave Taht]

On Sat, Mar 29, 2014 at 12:27 PM, Martin Bailey <martin@pcalpha.com> wrote:
> The recently released v2 of the TP-Link Archer C7 router would be a
> pretty good low-cost replacement. It's fully open-source Atheros/Qualcomm,

Um, no, the ath10k depends on a binary blob. Which has been really
irksome. I do hope they open-source the firmware so we can make
more progress on incorporating better algorithms into wifi.

> includes a fairly high performance 720MHz SOC with 128MB RAM, 16MB
> flash and 6 dual-band antennas (AC1750) with very good range. It can
> be found for $99 right now. The first hardware revision isn't
> supported by the ath10k driver in OpenWRT so make sure to only
> consider v2.

It's not clear how to ensure you are buying a v2 through sites like amazon.

If you can find a good source for it let me know...

>
> http://wiki.openwrt.org/toh/tp-link/tl-wdr7500
> https://wikidevi.com/wiki/TP-LINK_Archer_C7_v2.x
> http://www.newegg.com/Product/Product.aspx?Item=N82E16833704177
>
>> On Thu, 27 Mar 2014, Aaron Wood wrote:
>>
>>> Date: Thu, 27 Mar 2014 13:50:27 +0100
>>> From: Aaron Wood <woody77 at gmail.com>
>>> To: David Lang <david at lang.hm>
>>> Cc: Dave Taht <dave.taht at gmail.com>,
>>>     "cerowrt-devel at lists.bufferbloat.net"
>>>     <cerowrt-devel at lists.bufferbloat.net>
>>> Subject: Re: [Cerowrt-devel] Fwd: wndr3800 replacement
>>>
>>> On Wed, Mar 26, 2014 at 11:11 PM, David Lang <david at lang.hm> wrote:
>>>
>>>> If the openwrt folks could figure out how they are going to deal with
>>>> NAND
>>>> flash, it would be nice to be able to use one of the many routers that is
>>>> shipping with more flash (128M in the newer netgear routers would be
>>>> nice)
>>>>
>>>> if I were to get my hands on one, what sort of testing would you want to
>>>> do to it to tell if it looks like it would hold up?
>>>
>>>
>>> I have experience running mtd on NAND, using jffs2.  It seems to be
>>> holding
>>> up well.  Better than NOR did, honestly.  Although in general, I wish they
>>> would shift to eMMC.  But it's driven by two factors:
>>>
>>> 1) part cost
>>> 2) chipset support from the router SoC vendors
>>>
>>> Given some of the wishes that I see on here, I think for development,
>>> people would be happier with a platform that wasn't based on a router SoC
>>> (like the wndr is), but instead was based on an embedded application
>>> processor with PCIe for the radios, and an external switch fabric.
>>
>> I think we have two competing desires.
>>
>> one is to have a nice powerful device for those people who have fast
>> connections
>> and for us to experiment with.
>>
>> the second is to have a 'home' device.
>>
>> using a 3800 or similarly priced ($100-$150 USD) device that's readily
>> available
>> is very good for the second category, the question is if we can find one
>> that's
>> powerful enough for the first.
>>
>> David Lang
>>
>>>  But for
>>> thermal purposes alone, I've been seeing more and more external switch
>>> fabrics.  The heat of a 5-port gigabit switch IC is pretty substantial
>>> (from my teardowns).
>>>
>>> One item I think will be a boon, especially with DNSSEC, is super-cap or
>>> battery-backed rtc, but that's asking for a unicorn, I think.  Or...  a
>>> Gateworks Ventana GW5310 loaded with a couple standard (industrial-grade)
>>> PCIe radios, loaded into a custom case.  My guess is that it's a pretty
>>> expensive route, though.  I would be surprised if a completely assembled
>>> unit would be <$300.  At which point it starts to look better to just run
>>> a
>>> separate router and AP (using standard wndr-type platforms as the APs and
>>> a
>>> higher-end board or PC as the gateway).
>>>
>>> -Aaron
>>>
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel



-- 
Dave Täht

Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Martin Bailey]

You're right, there is that proprietary firmware bit for 5GHz. It's
mostly but not fully open-source then, at least the firmware is easy
to access and update.

Newegg specifies v2 in the product description, that's the most
reliable source for now. I imagine Amazon will also run out of 1.0 and
1.1 inventory pretty soon if it's not already the case.

On Sat, Mar 29, 2014 at 12:56 PM, Dave Taht <dave.taht@gmail.com> wrote:
> On Sat, Mar 29, 2014 at 12:27 PM, Martin Bailey <martin@pcalpha.com> wrote:
>> The recently released v2 of the TP-Link Archer C7 router would be a
>> pretty good low-cost replacement. It's fully open-source Atheros/Qualcomm,
>
> Um, no, the ath10k depends on a binary blob. Which has been really
> irksome. I do hope they open-source the firmware so we can make
> more progress on incorporating better algorithms into wifi.
>
>> includes a fairly high performance 720MHz SOC with 128MB RAM, 16MB
>> flash and 6 dual-band antennas (AC1750) with very good range. It can
>> be found for $99 right now. The first hardware revision isn't
>> supported by the ath10k driver in OpenWRT so make sure to only
>> consider v2.
>
> It's not clear how to ensure you are buying a v2 through sites like amazon.
>
> If you can find a good source for it let me know...
>
>>
>> http://wiki.openwrt.org/toh/tp-link/tl-wdr7500
>> https://wikidevi.com/wiki/TP-LINK_Archer_C7_v2.x
>> http://www.newegg.com/Product/Product.aspx?Item=N82E16833704177
>>
>>> On Thu, 27 Mar 2014, Aaron Wood wrote:
>>>
>>>> Date: Thu, 27 Mar 2014 13:50:27 +0100
>>>> From: Aaron Wood <woody77 at gmail.com>
>>>> To: David Lang <david at lang.hm>
>>>> Cc: Dave Taht <dave.taht at gmail.com>,
>>>>     "cerowrt-devel at lists.bufferbloat.net"
>>>>     <cerowrt-devel at lists.bufferbloat.net>
>>>> Subject: Re: [Cerowrt-devel] Fwd: wndr3800 replacement
>>>>
>>>> On Wed, Mar 26, 2014 at 11:11 PM, David Lang <david at lang.hm> wrote:
>>>>
>>>>> If the openwrt folks could figure out how they are going to deal with
>>>>> NAND
>>>>> flash, it would be nice to be able to use one of the many routers that is
>>>>> shipping with more flash (128M in the newer netgear routers would be
>>>>> nice)
>>>>>
>>>>> if I were to get my hands on one, what sort of testing would you want to
>>>>> do to it to tell if it looks like it would hold up?
>>>>
>>>>
>>>> I have experience running mtd on NAND, using jffs2.  It seems to be
>>>> holding
>>>> up well.  Better than NOR did, honestly.  Although in general, I wish they
>>>> would shift to eMMC.  But it's driven by two factors:
>>>>
>>>> 1) part cost
>>>> 2) chipset support from the router SoC vendors
>>>>
>>>> Given some of the wishes that I see on here, I think for development,
>>>> people would be happier with a platform that wasn't based on a router SoC
>>>> (like the wndr is), but instead was based on an embedded application
>>>> processor with PCIe for the radios, and an external switch fabric.
>>>
>>> I think we have two competing desires.
>>>
>>> one is to have a nice powerful device for those people who have fast
>>> connections
>>> and for us to experiment with.
>>>
>>> the second is to have a 'home' device.
>>>
>>> using a 3800 or similarly priced ($100-$150 USD) device that's readily
>>> available
>>> is very good for the second category, the question is if we can find one
>>> that's
>>> powerful enough for the first.
>>>
>>> David Lang
>>>
>>>>  But for
>>>> thermal purposes alone, I've been seeing more and more external switch
>>>> fabrics.  The heat of a 5-port gigabit switch IC is pretty substantial
>>>> (from my teardowns).
>>>>
>>>> One item I think will be a boon, especially with DNSSEC, is super-cap or
>>>> battery-backed rtc, but that's asking for a unicorn, I think.  Or...  a
>>>> Gateworks Ventana GW5310 loaded with a couple standard (industrial-grade)
>>>> PCIe radios, loaded into a custom case.  My guess is that it's a pretty
>>>> expensive route, though.  I would be surprised if a completely assembled
>>>> unit would be <$300.  At which point it starts to look better to just run
>>>> a
>>>> separate router and AP (using standard wndr-type platforms as the APs and
>>>> a
>>>> higher-end board or PC as the gateway).
>>>>
>>>> -Aaron
>>>>
>> _______________________________________________
>> Cerowrt-devel mailing list
>> Cerowrt-devel@lists.bufferbloat.net
>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>
>
>
> --
> Dave Täht
>
> Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Dave Taht]

On Sun, Mar 30, 2014 at 3:03 PM, Michael Richardson <mcr@sandelman.ca> wrote:
>
> Dave Taht <dave.taht@gmail.com> wrote:
>     >> Dave Taht <dave.taht@gmail.com> wrote: >> that is just fine.  I don't
>     >> expect 24-ports of GbE.  >> (I was expecting that cisco switch to do
>     >> that... sadly no lost packets >> on a $100 unmanaged FE switch...)
>     >>
>     >> > I started looking at the edgerouter LTE and related boxes again.  >
>     >> Does anyone else have one?
>     >>
>     >> I have ordered the 5-port POE one.
>
>     > Does that support gigabit POE?
>
> Unclear...  web site says:
>             10/100/1000 Ethernet ports that support 24V or 48V PoE output
>             with software-selectable voltage control*
>
>     >> My understanding is that it runs Vyatta. I would suspect that given
>     >> that the
>
>     > It is running an older version of vyatta, yes.
>
> okay.
>
>     >> edgerouter is MIPS, and has "special hardware", that VyOS might not
>     >> have the
>
>     > Not sure what the offloads buy you, except added latency. They are
>     > easily disabled for testing various qdiscs... (can't remember the
>     > command offhand)
>
>     >> right support for that hardware.  I'm going to talk to Ubiquity if I
>     >> can on Monday.  It might not make sense to run *Wrt on this device,
>     >> not sure.
>
>     > There is a pretty modern looking up-to-date build of openwrt for the
>     > edgerouter, using 3.10, and so on. I am told however, that there are
>     > some problems with the toolchain, which show up when you try to use
>     > iptables. Haven't tried it myself (my edgerouter is 60km away), and
>     > don't know how to flash openwrt onto it in the first place.
>
> okay.
>
>     > the current edgerouter firmware (v.1.4.1) is based on 3.4.23, and has
>     > all the bugs in the 3.4 series.
>
>     > I backported fq_codel and the latest flow hashing stuff to that
>     > version; patch 0013 is a little problematic as yet:
>
> why did you backport, rather than move the kernel forward?

Getting a new kernel version stable is kind of a long, hard effort,
don't you think? Adding a single feature seems simpler...

> Are there custom pieces in 3.4 which need to be forward ported?

Doesn't look like it. Of course, any major kernel version change requires
extensive testing...

>     > http://snapon.lab.bufferbloat.net/~cero2/edgerouter-fq_codel-patches/

Wanted to convince ubnt that there was value in fq_codel. They only just
released v1.4.1 with 3.4 (a huge jump forward from their last
release), the prospect of another 3 year jump nonplussed them.

Besides the openwrt version is already at 3.10, and I was curious as to
being able to compare 3.4 vs 3.x, and how another OS implements QoS.
(they have a cisco IOS like interface that appears to be written in Perl)

I just retrieved my edgerouter this morning and have a kernel built...


>
> --
> ]               Never tell me the odds!                 | ipv6 mesh networks [
> ]   Michael Richardson, Sandelman Software Works        | network architect  [
> ]     mcr@sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [
>



-- 
Dave Täht

Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Michael Richardson]

Dave Taht <dave.taht@gmail.com> wrote:
    >> Dave Taht <dave.taht@gmail.com> wrote: >> that is just fine.  I don't
    >> expect 24-ports of GbE.  >> (I was expecting that cisco switch to do
    >> that... sadly no lost packets >> on a $100 unmanaged FE switch...)
    >> 
    >> > I started looking at the edgerouter LTE and related boxes again.  >
    >> Does anyone else have one?
    >> 
    >> I have ordered the 5-port POE one.

    > Does that support gigabit POE?

Unclear...  web site says:
            10/100/1000 Ethernet ports that support 24V or 48V PoE output
            with software-selectable voltage control* 

    >> My understanding is that it runs Vyatta. I would suspect that given
    >> that the

    > It is running an older version of vyatta, yes.

okay.

    >> edgerouter is MIPS, and has "special hardware", that VyOS might not
    >> have the

    > Not sure what the offloads buy you, except added latency. They are
    > easily disabled for testing various qdiscs... (can't remember the
    > command offhand)

    >> right support for that hardware.  I'm going to talk to Ubiquity if I
    >> can on Monday.  It might not make sense to run *Wrt on this device,
    >> not sure.

    > There is a pretty modern looking up-to-date build of openwrt for the
    > edgerouter, using 3.10, and so on. I am told however, that there are
    > some problems with the toolchain, which show up when you try to use
    > iptables. Haven't tried it myself (my edgerouter is 60km away), and
    > don't know how to flash openwrt onto it in the first place.

okay.

    > the current edgerouter firmware (v.1.4.1) is based on 3.4.23, and has
    > all the bugs in the 3.4 series.

    > I backported fq_codel and the latest flow hashing stuff to that
    > version; patch 0013 is a little problematic as yet:

why did you backport, rather than move the kernel forward?
Are there custom pieces in 3.4 which need to be forward ported?

    > http://snapon.lab.bufferbloat.net/~cero2/edgerouter-fq_codel-patches/

-- 
]               Never tell me the odds!                 | ipv6 mesh networks [ 
]   Michael Richardson, Sandelman Software Works        | network architect  [ 
]     mcr@sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [ 
	

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Dave Taht]

On Sat, Mar 29, 2014 at 2:08 PM, Michael Richardson <mcr@sandelman.ca> wrote:
>
> Dave Taht <dave.taht@gmail.com> wrote:
>     >> that is just fine.  I don't expect 24-ports of GbE.
>     >> (I was expecting that cisco switch to do that... sadly no lost packets
>     >> on a $100 unmanaged FE switch...)
>
>     > I started looking at the edgerouter LTE and related boxes again.
>     > Does anyone else have one?
>
> I have ordered the 5-port POE one.

Does that support gigabit POE?

> My understanding is that it runs Vyatta. I would suspect that given that the

It is running an older version of vyatta, yes.

> edgerouter is MIPS, and has "special hardware", that VyOS might not have the

Not sure what the offloads buy you, except added latency. They are
easily disabled for testing various qdiscs... (can't remember the
command offhand)

> right support for that hardware.  I'm going to talk to Ubiquity if I can on
> Monday.   It might not make sense to run *Wrt on this device, not sure.

There is a pretty modern looking up-to-date build of openwrt for the
edgerouter, using 3.10, and so on. I am told however, that there are
some problems with the toolchain, which show up when you try to use
iptables. Haven't tried it myself (my edgerouter is 60km away), and
don't know how to flash openwrt onto it in the first place.

the current edgerouter firmware (v.1.4.1) is based on 3.4.23, and has
all the bugs in the 3.4 series.

I backported fq_codel and the latest flow hashing stuff to that
version; patch 0013 is a little problematic as yet:

http://snapon.lab.bufferbloat.net/~cero2/edgerouter-fq_codel-patches/


>
> --
> ]               Never tell me the odds!                 | ipv6 mesh networks [
> ]   Michael Richardson, Sandelman Software Works        | network architect  [
> ]     mcr@sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [
>
>



-- 
Dave Täht

Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Michael Richardson]

Dave Taht <dave.taht@gmail.com> wrote:
    >> that is just fine.  I don't expect 24-ports of GbE.
    >> (I was expecting that cisco switch to do that... sadly no lost packets
    >> on a $100 unmanaged FE switch...)

    > I started looking at the edgerouter LTE and related boxes again.
    > Does anyone else have one?

I have ordered the 5-port POE one.

My understanding is that it runs Vyatta. I would suspect that given that the
edgerouter is MIPS, and has "special hardware", that VyOS might not have the
right support for that hardware.  I'm going to talk to Ubiquity if I can on
Monday.   It might not make sense to run *Wrt on this device, not sure.

--
]               Never tell me the odds!                 | ipv6 mesh networks [
]   Michael Richardson, Sandelman Software Works        | network architect  [
]     mcr@sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Aaron Wood]

[-- Attachment #1: Type: text/plain, Size: 1755 bytes --]

On Fri, Mar 28, 2014 at 8:39 PM, Dave Taht <dave.taht@gmail.com> wrote:

> On Fri, Mar 28, 2014 at 11:40 AM, Michael Richardson <mcr@sandelman.ca>
> wrote:
> >
> > Mikael Abrahamsson <swmike@swm.pp.se> wrote:
> >     > I also think that it'll be very hard to find a box that is decently
> >     > priced
> >     > that also will do gig speeds *and* will do AQM, since most home
> >     > gateways that
> >
> > So, I'm struggling to get a $700 Cisco (not Linksys) SOHO switch to
> forward
> > packets at GbE across my home LAN.   Gigabit ethernet in the home is
> actually
> > a reality.... and having a Gigabit LAN hit a 11Mb/s (half-duplex!)
> 802.11b
> > wireless means that in many homes, we actually need a box that satisfies
> > both.
>

Right, but in that instance, the flow-rates are only 802.11 rates, and a
wndr can keep up with that.  It's the sqm on the uplink/wan side that needs
the processing power.  Or, it's time to do what Dave has mentioned and
rewrite to get a faster implementation than htb.


> > My vote would be to double the price from $80 to $250, and set that as
> the
> > new base for cerowrt work.  If it can't forward more than 2-3 GbE links,
> > that is just fine.  I don't expect 24-ports of GbE.
> > (I was expecting that cisco switch to do that... sadly no lost packets
> > on a $100 unmanaged FE switch...)
>
>  I started looking at the edgerouter LTE and related boxes again.
>
> Does anyone else have one?
>
> They have a new release based on 3.4 out; backporting fq_codel might work.
>
> I don't have any numbers on it's performance however, and it does look like
> updating it to 3.10 would be good idea.
>

Nice looking platform.

Cavium Octeon?  Looks like it's all in software, and not in offload modules?

-Aaron

[-- Attachment #2: Type: text/html, Size: 2586 bytes --]

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Dave Taht]

On Fri, Mar 28, 2014 at 11:40 AM, Michael Richardson <mcr@sandelman.ca> wrote:
>
> Mikael Abrahamsson <swmike@swm.pp.se> wrote:
>     > I also think that it'll be very hard to find a box that is decently
>     > priced
>     > that also will do gig speeds *and* will do AQM, since most home
>     > gateways that
>
> So, I'm struggling to get a $700 Cisco (not Linksys) SOHO switch to forward
> packets at GbE across my home LAN.   Gigabit ethernet in the home is actually
> a reality.... and having a Gigabit LAN hit a 11Mb/s (half-duplex!) 802.11b
> wireless means that in many homes, we actually need a box that satisfies
> both.
>
> My vote would be to double the price from $80 to $250, and set that as the
> new base for cerowrt work.  If it can't forward more than 2-3 GbE links,
> that is just fine.  I don't expect 24-ports of GbE.
> (I was expecting that cisco switch to do that... sadly no lost packets
> on a $100 unmanaged FE switch...)

 I started looking at the edgerouter LTE and related boxes again.

Does anyone else have one?

They have a new release based on 3.4 out; backporting fq_codel might work.

I don't have any numbers on it's performance however, and it does look like
updating it to 3.10 would be good idea.

>
> --
> ]               Never tell me the odds!                 | ipv6 mesh networks [
> ]   Michael Richardson, Sandelman Software Works        | network architect  [
> ]     mcr@sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [
>
>
>
>
>
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel



-- 
Dave Täht

Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Michael Richardson]

BTW: I've been asked by a previous customer to look into adapting some
white label CPE device to run something with sane bufferbloat.
None of the Cisco/Netgear/etc. <$400 devices can protect a SIP/RTP
flow from a network upload.

A 3800-equivalent device would be very much ideal, but the packaging
of the current unit is too residential for their tastes. (rack mount would be
a nice option, but most clients won't have anything to screw it into).

Please unicast any suggestions you might have; people with products/skills
are also welcome.

--
]               Never tell me the odds!                 | ipv6 mesh networks [
]   Michael Richardson, Sandelman Software Works        | network architect  [
]     mcr@sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Michael Richardson]

Mikael Abrahamsson <swmike@swm.pp.se> wrote:
    > I also think that it'll be very hard to find a box that is decently
    > priced
    > that also will do gig speeds *and* will do AQM, since most home
    > gateways that

So, I'm struggling to get a $700 Cisco (not Linksys) SOHO switch to forward
packets at GbE across my home LAN.   Gigabit ethernet in the home is actually
a reality.... and having a Gigabit LAN hit a 11Mb/s (half-duplex!) 802.11b
wireless means that in many homes, we actually need a box that satisfies
both.

My vote would be to double the price from $80 to $250, and set that as the
new base for cerowrt work.  If it can't forward more than 2-3 GbE links,
that is just fine.  I don't expect 24-ports of GbE.
(I was expecting that cisco switch to do that... sadly no lost packets
on a $100 unmanaged FE switch...)

--
]               Never tell me the odds!                 | ipv6 mesh networks [
]   Michael Richardson, Sandelman Software Works        | network architect  [
]     mcr@sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Aaron Wood]

[-- Attachment #1: Type: text/plain, Size: 1796 bytes --]

On Fri, Mar 28, 2014 at 9:36 AM, Mikael Abrahamsson <swmike@swm.pp.se>wrote:

> On Thu, 27 Mar 2014, David Lang wrote:
>
>  using a 3800 or similarly priced ($100-$150 USD) device that's readily
>> available is very good for the second category, the question is if we can
>> find one that's powerful enough for the first.
>>
>
> I would say that it would be desireable to have a device that is available
> world-wide, that doesn't have a lot of different versions of the hardware
> that are different.
>
> I also think that it'll be very hard to find a box that is decently priced
> that also will do gig speeds *and* will do AQM, since most home gateways
> that are this quick has hardware acceleration that in turn requires vendor
> proprietary kernels to drive the hardware.
>
> So if you want to fulfil both, I think we need two platforms. A CPU based
> platform that is quick enough to do AQM in and CPU forwarding in gig speeds
> will most likely have to be Intel based and cost 300 USD or more.
>

Certainly difficult when you bring being a Wifi AP into the mix.  This is
where a two-box solution might work better, running something like the wndr
as an internal router/AP, and using a separate box for the high-speed AQM
for the edge routing?  A dual-port GigE platform seems like it would be
easier to find (or build via mini-itx).

I'm nearly certain that something like a dual- or quad-codre i.MX6 at >1GHz
could do the work, but those just aren't available in a "home router"
package, which is the killer.  The fact that the home routers are commodity
products, and built to cost with a very narrow feature set makes it
exceedingly difficult to locate platforms that are viable for
experimentation.

Basically, we need a high-performance-networking version of a Raspberry Pi.

-Aaron

[-- Attachment #2: Type: text/html, Size: 2367 bytes --]

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Sebastian Moeller]

Hi Mikael, hi List,


On Mar 28, 2014, at 09:36 , Mikael Abrahamsson <swmike@swm.pp.se> wrote:

> On Thu, 27 Mar 2014, David Lang wrote:
> 
>> using a 3800 or similarly priced ($100-$150 USD) device that's readily available is very good for the second category, the question is if we can find one that's powerful enough for the first.
> 
> I would say that it would be desireable to have a device that is available world-wide, that doesn't have a lot of different versions of the hardware that are different.
> 
> I also think that it'll be very hard to find a box that is decently priced that also will do gig speeds *and* will do AQM, since most home gateways that are this quick has hardware acceleration that in turn requires vendor proprietary kernels to drive the hardware.
> 
> So if you want to fulfil both, I think we need two platforms. A CPU based platform that is quick enough to do AQM in and CPU forwarding in gig speeds will most likely have to be Intel based and cost 300 USD or more.

	I am not sure whether a dual core 1.4GHz has well based cerleron (http://www.tigerdirect.com/applications/searchtools/item-details.asp?EdpNo=8919600) will be sufficient (in combination with a USB3 ethernet dongle to make up the fact that there is only one port built-in), but $179 seems like an acceptable price point, given 2GB ram and 16GB SSD. The cpu is a bit weaker than the one in Dave's NUC, 1.4 vs 1.7GHz, but $180 beats $310+ easily ;).
	I wonder what it would take to install cerowrt on this and make it run well.

Best Regards
	sebastian

> 
> I would be happen to be wrong on the last point though...
> 
> -- 
> Mikael Abrahamsson    email: swmike@swm.pp.se
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Mikael Abrahamsson]

On Thu, 27 Mar 2014, David Lang wrote:

> using a 3800 or similarly priced ($100-$150 USD) device that's readily 
> available is very good for the second category, the question is if we 
> can find one that's powerful enough for the first.

I would say that it would be desireable to have a device that is available 
world-wide, that doesn't have a lot of different versions of the hardware 
that are different.

I also think that it'll be very hard to find a box that is decently priced 
that also will do gig speeds *and* will do AQM, since most home gateways 
that are this quick has hardware acceleration that in turn requires vendor 
proprietary kernels to drive the hardware.

So if you want to fulfil both, I think we need two platforms. A CPU based 
platform that is quick enough to do AQM in and CPU forwarding in gig 
speeds will most likely have to be Intel based and cost 300 USD or more.

I would be happen to be wrong on the last point though...

-- 
Mikael Abrahamsson    email: swmike@swm.pp.se

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[David Lang]

On Thu, 27 Mar 2014, Aaron Wood wrote:

> Date: Thu, 27 Mar 2014 13:50:27 +0100
> From: Aaron Wood <woody77@gmail.com>
> To: David Lang <david@lang.hm>
> Cc: Dave Taht <dave.taht@gmail.com>,
>     "cerowrt-devel@lists.bufferbloat.net"
>     <cerowrt-devel@lists.bufferbloat.net>
> Subject: Re: [Cerowrt-devel] Fwd: wndr3800 replacement
> 
> On Wed, Mar 26, 2014 at 11:11 PM, David Lang <david@lang.hm> wrote:
>
>> If the openwrt folks could figure out how they are going to deal with NAND
>> flash, it would be nice to be able to use one of the many routers that is
>> shipping with more flash (128M in the newer netgear routers would be nice)
>>
>> if I were to get my hands on one, what sort of testing would you want to
>> do to it to tell if it looks like it would hold up?
>
>
> I have experience running mtd on NAND, using jffs2.  It seems to be holding
> up well.  Better than NOR did, honestly.  Although in general, I wish they
> would shift to eMMC.  But it's driven by two factors:
>
> 1) part cost
> 2) chipset support from the router SoC vendors
>
> Given some of the wishes that I see on here, I think for development,
> people would be happier with a platform that wasn't based on a router SoC
> (like the wndr is), but instead was based on an embedded application
> processor with PCIe for the radios, and an external switch fabric.

I think we have two competing desires.

one is to have a nice powerful device for those people who have fast connections 
and for us to experiment with.

the second is to have a 'home' device.

using a 3800 or similarly priced ($100-$150 USD) device that's readily available 
is very good for the second category, the question is if we can find one that's 
powerful enough for the first.

David Lang

>  But for
> thermal purposes alone, I've been seeing more and more external switch
> fabrics.  The heat of a 5-port gigabit switch IC is pretty substantial
> (from my teardowns).
>
> One item I think will be a boon, especially with DNSSEC, is super-cap or
> battery-backed rtc, but that's asking for a unicorn, I think.  Or...  a
> Gateworks Ventana GW5310 loaded with a couple standard (industrial-grade)
> PCIe radios, loaded into a custom case.  My guess is that it's a pretty
> expensive route, though.  I would be surprised if a completely assembled
> unit would be <$300.  At which point it starts to look better to just run a
> separate router and AP (using standard wndr-type platforms as the APs and a
> higher-end board or PC as the gateway).
>
> -Aaron
>

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[Aaron Wood]

[-- Attachment #1: Type: text/plain, Size: 1716 bytes --]

On Wed, Mar 26, 2014 at 11:11 PM, David Lang <david@lang.hm> wrote:

> If the openwrt folks could figure out how they are going to deal with NAND
> flash, it would be nice to be able to use one of the many routers that is
> shipping with more flash (128M in the newer netgear routers would be nice)
>
> if I were to get my hands on one, what sort of testing would you want to
> do to it to tell if it looks like it would hold up?


I have experience running mtd on NAND, using jffs2.  It seems to be holding
up well.  Better than NOR did, honestly.  Although in general, I wish they
would shift to eMMC.  But it's driven by two factors:

1) part cost
2) chipset support from the router SoC vendors

Given some of the wishes that I see on here, I think for development,
people would be happier with a platform that wasn't based on a router SoC
(like the wndr is), but instead was based on an embedded application
processor with PCIe for the radios, and an external switch fabric.  But for
thermal purposes alone, I've been seeing more and more external switch
fabrics.  The heat of a 5-port gigabit switch IC is pretty substantial
(from my teardowns).

One item I think will be a boon, especially with DNSSEC, is super-cap or
battery-backed rtc, but that's asking for a unicorn, I think.  Or...  a
Gateworks Ventana GW5310 loaded with a couple standard (industrial-grade)
PCIe radios, loaded into a custom case.  My guess is that it's a pretty
expensive route, though.  I would be surprised if a completely assembled
unit would be <$300.  At which point it starts to look better to just run a
separate router and AP (using standard wndr-type platforms as the APs and a
higher-end board or PC as the gateway).

-Aaron

[-- Attachment #2: Type: text/html, Size: 2158 bytes --]

Re: [Cerowrt-devel] Fwd: wndr3800 replacement

[David Lang]

One item that has been very nice about the 3800 is the unbricking is so 
bulletproof. this looks like it has a similar capability.

also, the amazon review talks about it having a Freescale P1014 cpu

If the openwrt folks could figure out how they are going to deal with NAND 
flash, it would be nice to be able to use one of the many routers that is 
shipping with more flash (128M in the newer netgear routers would be nice)

if I were to get my hands on one, what sort of testing would you want to do to 
it to tell if it looks like it would hold up?

David Lang

On Tue, 25 Mar 2014, Dave Taht wrote:

> From: Mikael Abrahamsson
>
>
> What do you think of http://wiki.openwrt.org/toh/tp-link/tl-wdr4900 as
> wndr3800 replacement?
>
> --
> Mikael Abrahamsson    email: swmike@swm.pp.se
>
>
> + seems to have decent openwrt support, has anyone tried one?
>
> +/- I used to like powerpc. Then apple got off it. It seems like arm
> has most of the momentum nowadays.
> +/- I am mostly in search of something that can honestly hit gigE
> forwarding rates, and run htb
> at 500Mbits/sec. Don't know if ANYTHING has that amount of oomph besides x86.
> - I care about build quality and temp range for a long lasting device.
> The wndr3800 was flown to 120k feet,
> for example and I've seen it running in 100F+ weather. tp-link does
> not strike me as devices that do that.
>
> Other opinions welcomed.
>
>

[Cerowrt-devel] Fwd: wndr3800 replacement

[Dave Taht]

---------- Forwarded message ----------
From: Mikael Abrahamsson


What do you think of http://wiki.openwrt.org/toh/tp-link/tl-wdr4900 as
wndr3800 replacement?

--
Mikael Abrahamsson    email: swmike@swm.pp.se


+ seems to have decent openwrt support, has anyone tried one?

+/- I used to like powerpc. Then apple got off it. It seems like arm
has most of the momentum nowadays.
+/- I am mostly in search of something that can honestly hit gigE
forwarding rates, and run htb
at 500Mbits/sec. Don't know if ANYTHING has that amount of oomph besides x86.
- I care about build quality and temp range for a long lasting device.
The wndr3800 was flown to 120k feet,
for example and I've seen it running in 100F+ weather. tp-link does
not strike me as devices that do that.

Other opinions welcomed.

-- 
Dave Täht

Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html