[Cerowrt-devel] a smart SFP

[Cerowrt-devel] a smart SFP

[Dave Taht]

running linux, of course.

https://blog.benjojo.co.uk/post/smart-sfp-linux-inside

-- 
I tried to build a better future, a few times:
https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org

Dave Täht CEO, TekLibre, LLC

Re: [Cerowrt-devel] a smart SFP

[Sebastian Moeller]

That is similar to what happens in some GPON-ONT SFPs, some run a full small Linux distribution like OpenWrt inside.... though for ethernet that is unexpected.
This is also similar to SFP VDSL "modems" which likely run their own embedded OS as well inside the SFP package (at a time there was even a PCI VDSL2 "modem" that was actually running its own embedded system on the PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).

Regards
	Sebastian



> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
> 
> running linux, of course.
> 
> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
> 
> -- 
> I tried to build a better future, a few times:
> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> 
> Dave Täht CEO, TekLibre, LLC
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel

Re: [Cerowrt-devel] a smart SFP

[Dave Taht]

And a gpon onu

https://www.fs.com/products/133619.html

On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de> wrote:
>
> That is similar to what happens in some GPON-ONT SFPs, some run a full small Linux distribution like OpenWrt inside.... though for ethernet that is unexpected.
> This is also similar to SFP VDSL "modems" which likely run their own embedded OS as well inside the SFP package (at a time there was even a PCI VDSL2 "modem" that was actually running its own embedded system on the PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).
>
> Regards
>         Sebastian
>
>
>
> > On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
> >
> > running linux, of course.
> >
> > https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
> >
> > --
> > I tried to build a better future, a few times:
> > https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> >
> > Dave Täht CEO, TekLibre, LLC
> > _______________________________________________
> > Cerowrt-devel mailing list
> > Cerowrt-devel@lists.bufferbloat.net
> > https://lists.bufferbloat.net/listinfo/cerowrt-devel
>


-- 
I tried to build a better future, a few times:
https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org

Dave Täht CEO, TekLibre, LLC

Re: [Cerowrt-devel] a smart SFP

[Sebastian Moeller]

Yes sorry,

in GPON you have the unit at the telco side, often called OLT (optical line termination/terminal or similar, which feeds the whole segment of up to ~128 remote terminals and receives data from all) the passive splitter (if any) and the remote customer units that are called either ONU or ONT, but that are just two names for the same unit (not sure who uses which nomenclature, the local incumbent seems to prefer ONU, but their marketing name is Glasfasermodem, which just translates to fiber-modem). Nomenclature seems ot be consistent for the whole PON family, so in XG-PON or XGS-PON the unit apparently still are called OLT and ONU/ONT. And given that PON requires some smarts and configurability of the ONUs these tend to be small computers in their own right with their own little (or not so little) OSes.

Regards
	Sebastian



> On Jan 13, 2022, at 15:38, Dave Taht <dave.taht@gmail.com> wrote:
> 
> And a gpon onu
> 
> https://www.fs.com/products/133619.html
> 
> On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de> wrote:
>> 
>> That is similar to what happens in some GPON-ONT SFPs, some run a full small Linux distribution like OpenWrt inside.... though for ethernet that is unexpected.
>> This is also similar to SFP VDSL "modems" which likely run their own embedded OS as well inside the SFP package (at a time there was even a PCI VDSL2 "modem" that was actually running its own embedded system on the PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).
>> 
>> Regards
>>        Sebastian
>> 
>> 
>> 
>>> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
>>> 
>>> running linux, of course.
>>> 
>>> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
>>> 
>>> --
>>> I tried to build a better future, a few times:
>>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
>>> 
>>> Dave Täht CEO, TekLibre, LLC
>>> _______________________________________________
>>> Cerowrt-devel mailing list
>>> Cerowrt-devel@lists.bufferbloat.net
>>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>> 
> 
> 
> -- 
> I tried to build a better future, a few times:
> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> 
> Dave Täht CEO, TekLibre, LLC

Re: [Cerowrt-devel] a smart SFP

[Dave Taht]

yes, well, I dislike intensely the implications of
not-so-intelligently splitting those 128 terminals is, and was pleased
to see there was support
at least, for hardware flow control on the uplink side in the spec.
Since we've fixed dsl, cable, and wifi, getting gpon more right has
increasingly
been on my mind... so pulling  a testbed together of some sort would
be cool, and for that matter, having a SFP that could go right into a
SFP enabled home router rather than a separate unit seems like a good
idea, also.


On Thu, Jan 13, 2022 at 6:51 AM Sebastian Moeller <moeller0@gmx.de> wrote:
>
> Yes sorry,
>
> in GPON you have the unit at the telco side, often called OLT (optical line termination/terminal or similar, which feeds the whole segment of up to ~128 remote terminals and receives data from all) the passive splitter (if any) and the remote customer units that are called either ONU or ONT, but that are just two names for the same unit (not sure who uses which nomenclature, the local incumbent seems to prefer ONU, but their marketing name is Glasfasermodem, which just translates to fiber-modem). Nomenclature seems ot be consistent for the whole PON family, so in XG-PON or XGS-PON the unit apparently still are called OLT and ONU/ONT. And given that PON requires some smarts and configurability of the ONUs these tend to be small computers in their own right with their own little (or not so little) OSes.
>
> Regards
>         Sebastian
>
>
>
> > On Jan 13, 2022, at 15:38, Dave Taht <dave.taht@gmail.com> wrote:
> >
> > And a gpon onu
> >
> > https://www.fs.com/products/133619.html
> >
> > On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de> wrote:
> >>
> >> That is similar to what happens in some GPON-ONT SFPs, some run a full small Linux distribution like OpenWrt inside.... though for ethernet that is unexpected.
> >> This is also similar to SFP VDSL "modems" which likely run their own embedded OS as well inside the SFP package (at a time there was even a PCI VDSL2 "modem" that was actually running its own embedded system on the PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).
> >>
> >> Regards
> >>        Sebastian
> >>
> >>
> >>
> >>> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
> >>>
> >>> running linux, of course.
> >>>
> >>> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
> >>>
> >>> --
> >>> I tried to build a better future, a few times:
> >>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> >>>
> >>> Dave Täht CEO, TekLibre, LLC
> >>> _______________________________________________
> >>> Cerowrt-devel mailing list
> >>> Cerowrt-devel@lists.bufferbloat.net
> >>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
> >>
> >
> >
> > --
> > I tried to build a better future, a few times:
> > https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> >
> > Dave Täht CEO, TekLibre, LLC
>


-- 
I tried to build a better future, a few times:
https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org

Dave Täht CEO, TekLibre, LLC

Re: [Cerowrt-devel] a smart SFP

[Sebastian Moeller]

Hi Dave,


> On Jan 13, 2022, at 16:28, Dave Taht <dave.taht@gmail.com> wrote:
> 
> yes, well, I dislike intensely the implications of
> not-so-intelligently splitting those 128 terminals is, and was pleased
> to see there was support

	Nobody in their right mind is going to hook up 128 terminalt to one OLT port, I hope...


> at least, for hardware flow control on the uplink side in the spec.

	The spec, as many ITU documents contains probably more than anybody is going to use ;)
But GPON uses a request-grant system for upstream scheduling similar to DOCSIS (when looking at both from great height), just with a much faster message clock and hence less total and especially less variability in delay, if I understood things correctly.


> Since we've fixed dsl, cable, and wifi, getting gpon more right has
> increasingly
> been on my mind...

	Fist question might to be "how broken is GPON/XGPON" to start with, no?


> so pulling  a testbed together of some sort would
> be cool, and for that matter, having a SFP that could go right into a
> SFP enabled home router rather than a separate unit seems like a good
> idea, also.

	Meh, unless you actually get root on the ONU, it will basically act like a small media converter whether you connect via SFP or via ethernet. In Germany enthusiasts started to look at SFP-ONUs partly because ISPs were provisioning more GPON "rate" than required for that the ~940 Mbps goodput of gigabit ethernet, going SFP allowed some to run (compatible SFP-onus) at 2.5 Gbps link rate to the SFP cage hence allowing speedtest superiority... (that and to clone ONUs so you can have a replacement in the cupboard in case your original dies; typicallY ONUs need to be individually provisioned by the ISP but if you have a perfect clone all the ISP would see is a short power-down). The upshot of this is that some found ways onto SFP onus so you can get more information from the ONU if you are really dedicated. But first you need an GPON link and an ISP that will provision whatever ONU you selected... (ITU envisioned the ONU to be part of the ISPs network and hence little thought was spend on ideas like giving the end user root on those devices). And that is where I am out, currently FTTH-built out is a big topic at home, but has not reached my flat yet and might not for a few years (oh, the irony, as I am living literally next door to a central office of the incumbent telco, spanning a fiber cable over less then 20 m should get me FTTH, but I digress)


Regards
	Sebastian


> 
> 
> On Thu, Jan 13, 2022 at 6:51 AM Sebastian Moeller <moeller0@gmx.de> wrote:
>> 
>> Yes sorry,
>> 
>> in GPON you have the unit at the telco side, often called OLT (optical line termination/terminal or similar, which feeds the whole segment of up to ~128 remote terminals and receives data from all) the passive splitter (if any) and the remote customer units that are called either ONU or ONT, but that are just two names for the same unit (not sure who uses which nomenclature, the local incumbent seems to prefer ONU, but their marketing name is Glasfasermodem, which just translates to fiber-modem). Nomenclature seems ot be consistent for the whole PON family, so in XG-PON or XGS-PON the unit apparently still are called OLT and ONU/ONT. And given that PON requires some smarts and configurability of the ONUs these tend to be small computers in their own right with their own little (or not so little) OSes.
>> 
>> Regards
>>        Sebastian
>> 
>> 
>> 
>>> On Jan 13, 2022, at 15:38, Dave Taht <dave.taht@gmail.com> wrote:
>>> 
>>> And a gpon onu
>>> 
>>> https://www.fs.com/products/133619.html
>>> 
>>> On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de> wrote:
>>>> 
>>>> That is similar to what happens in some GPON-ONT SFPs, some run a full small Linux distribution like OpenWrt inside.... though for ethernet that is unexpected.
>>>> This is also similar to SFP VDSL "modems" which likely run their own embedded OS as well inside the SFP package (at a time there was even a PCI VDSL2 "modem" that was actually running its own embedded system on the PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).
>>>> 
>>>> Regards
>>>>       Sebastian
>>>> 
>>>> 
>>>> 
>>>>> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
>>>>> 
>>>>> running linux, of course.
>>>>> 
>>>>> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
>>>>> 
>>>>> --
>>>>> I tried to build a better future, a few times:
>>>>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
>>>>> 
>>>>> Dave Täht CEO, TekLibre, LLC
>>>>> _______________________________________________
>>>>> Cerowrt-devel mailing list
>>>>> Cerowrt-devel@lists.bufferbloat.net
>>>>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>>>> 
>>> 
>>> 
>>> --
>>> I tried to build a better future, a few times:
>>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
>>> 
>>> Dave Täht CEO, TekLibre, LLC
>> 
> 
> 
> -- 
> I tried to build a better future, a few times:
> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> 
> Dave Täht CEO, TekLibre, LLC

Re: [Cerowrt-devel] a smart SFP

[Sebastian Moeller]

Hi Dave,


this thread https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/ (in German) has some instructions how to get root on one type of SFP ONU... (I was monitoring that thread for general interest, turns out the intel falcon plattform seems somehow based on an ancient OpenWrt)

Regards
	Sebastian


> On Jan 13, 2022, at 15:38, Dave Taht <dave.taht@gmail.com> wrote:
> 
> And a gpon onu
> 
> https://www.fs.com/products/133619.html
> 
> On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de> wrote:
>> 
>> That is similar to what happens in some GPON-ONT SFPs, some run a full small Linux distribution like OpenWrt inside.... though for ethernet that is unexpected.
>> This is also similar to SFP VDSL "modems" which likely run their own embedded OS as well inside the SFP package (at a time there was even a PCI VDSL2 "modem" that was actually running its own embedded system on the PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).
>> 
>> Regards
>>        Sebastian
>> 
>> 
>> 
>>> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
>>> 
>>> running linux, of course.
>>> 
>>> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
>>> 
>>> --
>>> I tried to build a better future, a few times:
>>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
>>> 
>>> Dave Täht CEO, TekLibre, LLC
>>> _______________________________________________
>>> Cerowrt-devel mailing list
>>> Cerowrt-devel@lists.bufferbloat.net
>>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>> 
> 
> 
> -- 
> I tried to build a better future, a few times:
> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> 
> Dave Täht CEO, TekLibre, LLC

Re: [Cerowrt-devel] a smart SFP

[Dave Taht]

On Thu, Jan 13, 2022 at 7:57 AM Sebastian Moeller <moeller0@gmx.de> wrote:
>
> Hi Dave,
>
>
> this thread https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/ (in German) has some instructions how to get root on one type of SFP ONU... (I was monitoring that thread for general interest, turns out the intel falcon plattform seems somehow based on an ancient OpenWrt)
>
> Regards
>         Sebastian

It's really remarkable how many places are running an ancient openwrt.
Starlink's use was not an abomination, but a persistent reality. Given
how much
chaos calmer I've found, I sometimes wish we'd somehow started the
cerowrt project 2 years earlier. Then we'd be done by now.

>
> > On Jan 13, 2022, at 15:38, Dave Taht <dave.taht@gmail.com> wrote:
> >
> > And a gpon onu
> >
> > https://www.fs.com/products/133619.html
> >
> > On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de> wrote:
> >>
> >> That is similar to what happens in some GPON-ONT SFPs, some run a full small Linux distribution like OpenWrt inside.... though for ethernet that is unexpected.
> >> This is also similar to SFP VDSL "modems" which likely run their own embedded OS as well inside the SFP package (at a time there was even a PCI VDSL2 "modem" that was actually running its own embedded system on the PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).
> >>
> >> Regards
> >>        Sebastian
> >>
> >>
> >>
> >>> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
> >>>
> >>> running linux, of course.
> >>>
> >>> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
> >>>
> >>> --
> >>> I tried to build a better future, a few times:
> >>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> >>>
> >>> Dave Täht CEO, TekLibre, LLC
> >>> _______________________________________________
> >>> Cerowrt-devel mailing list
> >>> Cerowrt-devel@lists.bufferbloat.net
> >>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
> >>
> >
> >
> > --
> > I tried to build a better future, a few times:
> > https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> >
> > Dave Täht CEO, TekLibre, LLC
>


-- 
I tried to build a better future, a few times:
https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org

Dave Täht CEO, TekLibre, LLC

Re: [Cerowrt-devel] a smart SFP

[Sebastian Moeller]

Hi Dave,



> On Jan 13, 2022, at 16:59, Dave Taht <dave.taht@gmail.com> wrote:
> 
> On Thu, Jan 13, 2022 at 7:57 AM Sebastian Moeller <moeller0@gmx.de> wrote:
>> 
>> Hi Dave,
>> 
>> 
>> this thread https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/ (in German) has some instructions how to get root on one type of SFP ONU... (I was monitoring that thread for general interest, turns out the intel falcon plattform seems somehow based on an ancient OpenWrt)
>> 
>> Regards
>>        Sebastian
> 
> It's really remarkable how many places are running an ancient openwrt.
> Starlink's use was not an abomination, but a persistent reality. Given
> how much
> chaos calmer I've found, I sometimes wish we'd somehow started the
> cerowrt project 2 years earlier.

	Yes and no.

> Then we'd be done by now.

	Hopefully, but then I would not have noticed the whole thing and would probably not have participated... ;)

Regards
	Sebastian 



> 
>> 
>>> On Jan 13, 2022, at 15:38, Dave Taht <dave.taht@gmail.com> wrote:
>>> 
>>> And a gpon onu
>>> 
>>> https://www.fs.com/products/133619.html
>>> 
>>> On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de> wrote:
>>>> 
>>>> That is similar to what happens in some GPON-ONT SFPs, some run a full small Linux distribution like OpenWrt inside.... though for ethernet that is unexpected.
>>>> This is also similar to SFP VDSL "modems" which likely run their own embedded OS as well inside the SFP package (at a time there was even a PCI VDSL2 "modem" that was actually running its own embedded system on the PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).
>>>> 
>>>> Regards
>>>>       Sebastian
>>>> 
>>>> 
>>>> 
>>>>> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
>>>>> 
>>>>> running linux, of course.
>>>>> 
>>>>> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
>>>>> 
>>>>> --
>>>>> I tried to build a better future, a few times:
>>>>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
>>>>> 
>>>>> Dave Täht CEO, TekLibre, LLC
>>>>> _______________________________________________
>>>>> Cerowrt-devel mailing list
>>>>> Cerowrt-devel@lists.bufferbloat.net
>>>>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>>>> 
>>> 
>>> 
>>> --
>>> I tried to build a better future, a few times:
>>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
>>> 
>>> Dave Täht CEO, TekLibre, LLC
>> 
> 
> 
> -- 
> I tried to build a better future, a few times:
> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> 
> Dave Täht CEO, TekLibre, LLC

Re: [Cerowrt-devel] a smart SFP

[Jonas Mårtensson]

[-- Attachment #1: Type: text/plain, Size: 4942 bytes --]

Hi,

>  getting gpon more right has increasingly been on my mind

I think more right is to not turn the fiber into a shared medium in the
first place but since gpon is so popular, improving it seems like a nice
goal.

> Nobody in their right mind is going to hook up 128 terminalt to one OLT
port, I hope...

Well, sharing one OLT port between many terminals is kind of the (only)
advantage of PON, although split ratios of 32 or 64 are more typical. But
often it's the loss budget that limits the ratio.

>  Fist question might to be "how broken is GPON/XGPON" to start with, no?

Looking at dslreports bufferbloat results for fiber, there are many samples
with >250ms latency on the uplink. Unfortunately, this graph doesn't show
results for 500Mbit/s or 1Gbit/s services but it's still interesting to
look at:

https://www.dslreports.com/speedtest/results/bufferbloat?up=1

>  this thread
https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/
(in
German) has some instructions how to get root on one type of SFP ONU...

Thanks, that's an interesting thread. "Hacking" SFP ONUs seems like a
popular hobby. Here are some other resources I found:

https://github.com/zry98/SFP-GPON-ONU
https://github.com/hwti/G-010S-A
https://forum.mikrotik.com/viewtopic.php?t=116364&start=300#p771961

> so pulling  a testbed together of some sort would be cool, and for that
matter, having a SFP that could go right into a SFP enabled home router
rather than a separate unit seems like a good idea, also

Yes, but ideally I guess you would also need some control of the OLT side.
You may want to look into the VOLTHA project run by ONF:

https://wiki.opennetworking.org/display/COM/VOLTHA

/Jonas

On Thu, Jan 13, 2022 at 5:29 PM Sebastian Moeller <moeller0@gmx.de> wrote:

> Hi Dave,
>
>
>
> > On Jan 13, 2022, at 16:59, Dave Taht <dave.taht@gmail.com> wrote:
> >
> > On Thu, Jan 13, 2022 at 7:57 AM Sebastian Moeller <moeller0@gmx.de>
> wrote:
> >>
> >> Hi Dave,
> >>
> >>
> >> this thread
> https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/
> (in German) has some instructions how to get root on one type of SFP ONU...
> (I was monitoring that thread for general interest, turns out the intel
> falcon plattform seems somehow based on an ancient OpenWrt)
> >>
> >> Regards
> >>        Sebastian
> >
> > It's really remarkable how many places are running an ancient openwrt.
> > Starlink's use was not an abomination, but a persistent reality. Given
> > how much
> > chaos calmer I've found, I sometimes wish we'd somehow started the
> > cerowrt project 2 years earlier.
>
>         Yes and no.
>
> > Then we'd be done by now.
>
>         Hopefully, but then I would not have noticed the whole thing and
> would probably not have participated... ;)
>
> Regards
>         Sebastian
>
>
>
> >
> >>
> >>> On Jan 13, 2022, at 15:38, Dave Taht <dave.taht@gmail.com> wrote:
> >>>
> >>> And a gpon onu
> >>>
> >>> https://www.fs.com/products/133619.html
> >>>
> >>> On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de>
> wrote:
> >>>>
> >>>> That is similar to what happens in some GPON-ONT SFPs, some run a
> full small Linux distribution like OpenWrt inside.... though for ethernet
> that is unexpected.
> >>>> This is also similar to SFP VDSL "modems" which likely run their own
> embedded OS as well inside the SFP package (at a time there was even a PCI
> VDSL2 "modem" that was actually running its own embedded system on the PCI
> board, IIRC, it pretended to the main computer to be an ethernet NIC).
> >>>>
> >>>> Regards
> >>>>       Sebastian
> >>>>
> >>>>
> >>>>
> >>>>> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
> >>>>>
> >>>>> running linux, of course.
> >>>>>
> >>>>> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
> >>>>>
> >>>>> --
> >>>>> I tried to build a better future, a few times:
> >>>>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> >>>>>
> >>>>> Dave Täht CEO, TekLibre, LLC
> >>>>> _______________________________________________
> >>>>> Cerowrt-devel mailing list
> >>>>> Cerowrt-devel@lists.bufferbloat.net
> >>>>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
> >>>>
> >>>
> >>>
> >>> --
> >>> I tried to build a better future, a few times:
> >>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> >>>
> >>> Dave Täht CEO, TekLibre, LLC
> >>
> >
> >
> > --
> > I tried to build a better future, a few times:
> > https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> >
> > Dave Täht CEO, TekLibre, LLC
>
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>

[-- Attachment #2: Type: text/html, Size: 8357 bytes --]

Re: [Cerowrt-devel] a smart SFP

[Sebastian Moeller]

Hi Jonas,


> On Jan 14, 2022, at 11:44, Jonas Mårtensson <martensson.jonas@gmail.com> wrote:
> 
> Hi,
> 
> >  getting gpon more right has increasingly been on my mind
> 
> I think more right is to not turn the fiber into a shared medium in the first place but since gpon is so popular, improving it seems like a nice goal.
> 
> > Nobody in their right mind is going to hook up 128 terminalt to one OLT port, I hope...
> 
> Well, sharing one OLT port between many terminals is kind of the (only) advantage of PON, although split ratios of 32 or 64 are more typical. But often it's the loss budget that limits the ratio.

	Sure, but given that you probably need a few splices along the way and preferably pluggable connectors at both ends the loss budget is not that large (assuming an ISP does not want to push its luck and allows for stuff like end-users not cleaning the plug diligently before each plugging).


> 
> >  Fist question might to be "how broken is GPON/XGPON" to start with, no?
> 
> Looking at dslreports bufferbloat results for fiber, there are many samples with >250ms latency on the uplink. Unfortunately, this graph doesn't show results for 500Mbit/s or 1Gbit/s services but it's still interesting to look at:
> 
> https://www.dslreports.com/speedtest/results/bufferbloat?up=1

	Dslreports has no cue what a link is actually using, all it reports wich test profile a user selected, and some users like me ignore the names and simply use/recommend the profile with the desired number of flows. Plus quite a number of dedidedly metallic access technology are marketed with fiber somewhere in the name, potentially confusing users into selecting the "wrong" profile (think Fiber to the Cabinet for copper DSL or even Hybrid-Fiber-Coax for docsis cable)... in short the abels are nice, but I would not read too much inside those.


> 
> >  this thread https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/ (in German) has some instructions how to get root on one type of SFP ONU...
> 
> Thanks, that's an interesting thread. "Hacking" SFP ONUs seems like a popular hobby. Here are some other resources I found:
> 
> https://github.com/zry98/SFP-GPON-ONU
> https://github.com/hwti/G-010S-A
> https://forum.mikrotik.com/viewtopic.php?t=116364&start=300#p771961

Thanks for the links!

Regards
	Sebastian





> 
> > so pulling  a testbed together of some sort would be cool, and for that matter, having a SFP that could go right into a SFP enabled home router rather than a separate unit seems like a good idea, also
> 
> Yes, but ideally I guess you would also need some control of the OLT side. You may want to look into the VOLTHA project run by ONF:
> 
> https://wiki.opennetworking.org/display/COM/VOLTHA
> 
> /Jonas
> 
> On Thu, Jan 13, 2022 at 5:29 PM Sebastian Moeller <moeller0@gmx.de> wrote:
> Hi Dave,
> 
> 
> 
> > On Jan 13, 2022, at 16:59, Dave Taht <dave.taht@gmail.com> wrote:
> > 
> > On Thu, Jan 13, 2022 at 7:57 AM Sebastian Moeller <moeller0@gmx.de> wrote:
> >> 
> >> Hi Dave,
> >> 
> >> 
> >> this thread https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/ (in German) has some instructions how to get root on one type of SFP ONU... (I was monitoring that thread for general interest, turns out the intel falcon plattform seems somehow based on an ancient OpenWrt)
> >> 
> >> Regards
> >>        Sebastian
> > 
> > It's really remarkable how many places are running an ancient openwrt.
> > Starlink's use was not an abomination, but a persistent reality. Given
> > how much
> > chaos calmer I've found, I sometimes wish we'd somehow started the
> > cerowrt project 2 years earlier.
> 
>         Yes and no.
> 
> > Then we'd be done by now.
> 
>         Hopefully, but then I would not have noticed the whole thing and would probably not have participated... ;)
> 
> Regards
>         Sebastian 
> 
> 
> 
> > 
> >> 
> >>> On Jan 13, 2022, at 15:38, Dave Taht <dave.taht@gmail.com> wrote:
> >>> 
> >>> And a gpon onu
> >>> 
> >>> https://www.fs.com/products/133619.html
> >>> 
> >>> On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de> wrote:
> >>>> 
> >>>> That is similar to what happens in some GPON-ONT SFPs, some run a full small Linux distribution like OpenWrt inside.... though for ethernet that is unexpected.
> >>>> This is also similar to SFP VDSL "modems" which likely run their own embedded OS as well inside the SFP package (at a time there was even a PCI VDSL2 "modem" that was actually running its own embedded system on the PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).
> >>>> 
> >>>> Regards
> >>>>       Sebastian
> >>>> 
> >>>> 
> >>>> 
> >>>>> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
> >>>>> 
> >>>>> running linux, of course.
> >>>>> 
> >>>>> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
> >>>>> 
> >>>>> --
> >>>>> I tried to build a better future, a few times:
> >>>>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> >>>>> 
> >>>>> Dave Täht CEO, TekLibre, LLC
> >>>>> _______________________________________________
> >>>>> Cerowrt-devel mailing list
> >>>>> Cerowrt-devel@lists.bufferbloat.net
> >>>>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
> >>>> 
> >>> 
> >>> 
> >>> --
> >>> I tried to build a better future, a few times:
> >>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> >>> 
> >>> Dave Täht CEO, TekLibre, LLC
> >> 
> > 
> > 
> > -- 
> > I tried to build a better future, a few times:
> > https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> > 
> > Dave Täht CEO, TekLibre, LLC
> 
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel

Re: [Cerowrt-devel] a smart SFP

[Jonas Mårtensson]

[-- Attachment #1: Type: text/plain, Size: 7991 bytes --]

>  Sure, but given that you probably need a few splices along the way and
preferably pluggable connectors at both ends the loss budget is not that
large (assuming an ISP does not want to push its luck and allows for stuff
like end-users not cleaning the plug diligently before each plugging).

GPON loss budget is 28 dB and typical insertion losses for 1:32 and 1:64
splitters are 17 dB and 21 dB. This leaves 11 dB or 7 dB for splices,
connectors and fiber losses. I don't think it's common to have end-users
cleaning and plugging in the fiber, this is done by the ISP technician at
installation.

> Dslreports has no cue what a link is actually using, all it reports wich
test profile a user selected, and some users like me ignore the names and
simply use/recommend the profile with the desired number of flows. Plus
quite a number of dedidedly metallic access technology are marketed with
fiber somewhere in the name, potentially confusing users into selecting the
"wrong" profile (think Fiber to the Cabinet for copper DSL or even
Hybrid-Fiber-Coax for docsis cable)... in short the abels are nice, but I
would not read too much inside those.

I agree with all these points. It may be better to look at ISPs that are
known to only use PON, such as Google Fiber. Here are some recent tests
that all show similar and interesting bufferbloat behaviour on the uplink:

https://www.dslreports.com/speedtest/70320015
https://www.dslreports.com/speedtest/70346586
https://www.dslreports.com/speedtest/70346578

/Jonas

On Fri, Jan 14, 2022 at 12:55 PM Sebastian Moeller <moeller0@gmx.de> wrote:

> Hi Jonas,
>
>
> > On Jan 14, 2022, at 11:44, Jonas Mårtensson <martensson.jonas@gmail.com>
> wrote:
> >
> > Hi,
> >
> > >  getting gpon more right has increasingly been on my mind
> >
> > I think more right is to not turn the fiber into a shared medium in the
> first place but since gpon is so popular, improving it seems like a nice
> goal.
> >
> > > Nobody in their right mind is going to hook up 128 terminalt to one
> OLT port, I hope...
> >
> > Well, sharing one OLT port between many terminals is kind of the (only)
> advantage of PON, although split ratios of 32 or 64 are more typical. But
> often it's the loss budget that limits the ratio.
>
>         Sure, but given that you probably need a few splices along the way
> and preferably pluggable connectors at both ends the loss budget is not
> that large (assuming an ISP does not want to push its luck and allows for
> stuff like end-users not cleaning the plug diligently before each plugging).
>
>
> >
> > >  Fist question might to be "how broken is GPON/XGPON" to start with,
> no?
> >
> > Looking at dslreports bufferbloat results for fiber, there are many
> samples with >250ms latency on the uplink. Unfortunately, this graph
> doesn't show results for 500Mbit/s or 1Gbit/s services but it's still
> interesting to look at:
> >
> > https://www.dslreports.com/speedtest/results/bufferbloat?up=1
>
>         Dslreports has no cue what a link is actually using, all it
> reports wich test profile a user selected, and some users like me ignore
> the names and simply use/recommend the profile with the desired number of
> flows. Plus quite a number of dedidedly metallic access technology are
> marketed with fiber somewhere in the name, potentially confusing users into
> selecting the "wrong" profile (think Fiber to the Cabinet for copper DSL or
> even Hybrid-Fiber-Coax for docsis cable)... in short the abels are nice,
> but I would not read too much inside those.
>
>
> >
> > >  this thread
> https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/
> (in German) has some instructions how to get root on one type of SFP ONU...
> >
> > Thanks, that's an interesting thread. "Hacking" SFP ONUs seems like a
> popular hobby. Here are some other resources I found:
> >
> > https://github.com/zry98/SFP-GPON-ONU
> > https://github.com/hwti/G-010S-A
> > https://forum.mikrotik.com/viewtopic.php?t=116364&start=300#p771961
>
> Thanks for the links!
>
> Regards
>         Sebastian
>
>
>
>
>
> >
> > > so pulling  a testbed together of some sort would be cool, and for
> that matter, having a SFP that could go right into a SFP enabled home
> router rather than a separate unit seems like a good idea, also
> >
> > Yes, but ideally I guess you would also need some control of the OLT
> side. You may want to look into the VOLTHA project run by ONF:
> >
> > https://wiki.opennetworking.org/display/COM/VOLTHA
> >
> > /Jonas
> >
> > On Thu, Jan 13, 2022 at 5:29 PM Sebastian Moeller <moeller0@gmx.de>
> wrote:
> > Hi Dave,
> >
> >
> >
> > > On Jan 13, 2022, at 16:59, Dave Taht <dave.taht@gmail.com> wrote:
> > >
> > > On Thu, Jan 13, 2022 at 7:57 AM Sebastian Moeller <moeller0@gmx.de>
> wrote:
> > >>
> > >> Hi Dave,
> > >>
> > >>
> > >> this thread
> https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/
> (in German) has some instructions how to get root on one type of SFP ONU...
> (I was monitoring that thread for general interest, turns out the intel
> falcon plattform seems somehow based on an ancient OpenWrt)
> > >>
> > >> Regards
> > >>        Sebastian
> > >
> > > It's really remarkable how many places are running an ancient openwrt.
> > > Starlink's use was not an abomination, but a persistent reality. Given
> > > how much
> > > chaos calmer I've found, I sometimes wish we'd somehow started the
> > > cerowrt project 2 years earlier.
> >
> >         Yes and no.
> >
> > > Then we'd be done by now.
> >
> >         Hopefully, but then I would not have noticed the whole thing and
> would probably not have participated... ;)
> >
> > Regards
> >         Sebastian
> >
> >
> >
> > >
> > >>
> > >>> On Jan 13, 2022, at 15:38, Dave Taht <dave.taht@gmail.com> wrote:
> > >>>
> > >>> And a gpon onu
> > >>>
> > >>> https://www.fs.com/products/133619.html
> > >>>
> > >>> On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de>
> wrote:
> > >>>>
> > >>>> That is similar to what happens in some GPON-ONT SFPs, some run a
> full small Linux distribution like OpenWrt inside.... though for ethernet
> that is unexpected.
> > >>>> This is also similar to SFP VDSL "modems" which likely run their
> own embedded OS as well inside the SFP package (at a time there was even a
> PCI VDSL2 "modem" that was actually running its own embedded system on the
> PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).
> > >>>>
> > >>>> Regards
> > >>>>       Sebastian
> > >>>>
> > >>>>
> > >>>>
> > >>>>> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
> > >>>>>
> > >>>>> running linux, of course.
> > >>>>>
> > >>>>> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
> > >>>>>
> > >>>>> --
> > >>>>> I tried to build a better future, a few times:
> > >>>>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> > >>>>>
> > >>>>> Dave Täht CEO, TekLibre, LLC
> > >>>>> _______________________________________________
> > >>>>> Cerowrt-devel mailing list
> > >>>>> Cerowrt-devel@lists.bufferbloat.net
> > >>>>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
> > >>>>
> > >>>
> > >>>
> > >>> --
> > >>> I tried to build a better future, a few times:
> > >>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> > >>>
> > >>> Dave Täht CEO, TekLibre, LLC
> > >>
> > >
> > >
> > > --
> > > I tried to build a better future, a few times:
> > > https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> > >
> > > Dave Täht CEO, TekLibre, LLC
> >
> > _______________________________________________
> > Cerowrt-devel mailing list
> > Cerowrt-devel@lists.bufferbloat.net
> > https://lists.bufferbloat.net/listinfo/cerowrt-devel
>
>

[-- Attachment #2: Type: text/html, Size: 12025 bytes --]

Re: [Cerowrt-devel] a smart SFP

[Sebastian Moeller]

Hi Jonas,



> On Jan 14, 2022, at 14:12, Jonas Mårtensson <martensson.jonas@gmail.com> wrote:
> 
> >  Sure, but given that you probably need a few splices along the way and preferably pluggable connectors at both ends the loss budget is not that large (assuming an ISP does not want to push its luck and allows for stuff like end-users not cleaning the plug diligently before each plugging).
> 
> GPON loss budget is 28 dB and typical insertion losses for 1:32 and 1:64 splitters are 17 dB and 21 dB. This leaves 11 dB or 7 dB for splices, connectors and fiber losses. I don't think it's common to have end-users cleaning and plugging in the fiber, this is done by the ISP technician at installation.

	Not all ISPs work with a single level of splitters, some do 1:4 and split each of the four up into more later, I would guess that 1:4 + 1:8 will have a bigger aggregate attenuation than going 1:32 directly, but I am guessing here, so thanks for the numbers. In Germany the law is as it is and requires a passive handover point of an ISPs network to the home network and also the freedom to choose routers (and probably also ONTs/ONUs but that is still in flux), so plugs will be common as will be exchange of ONTs/ONUs by end-users. Whether that is a good or a bad thins is open for discussion, but as an ISP I would try to plan my PON plant such that there would be more loss reserve for these final connections than would be if these would be performed and documented by trained technicians. (I think that might be one of the consequences of deploying FTTH in the mass market, solutions need to be a bit more error tolerant than for networks mainly handled by experts only.)

> 
> > Dslreports has no cue what a link is actually using, all it reports wich test profile a user selected, and some users like me ignore the names and simply use/recommend the profile with the desired number of flows. Plus quite a number of dedidedly metallic access technology are marketed with fiber somewhere in the name, potentially confusing users into selecting the "wrong" profile (think Fiber to the Cabinet for copper DSL or even Hybrid-Fiber-Coax for docsis cable)... in short the abels are nice, but I would not read too much inside those.
> 
> I agree with all these points. It may be better to look at ISPs that are known to only use PON, such as Google Fiber. Here are some recent tests that all show similar and interesting bufferbloat behaviour on the uplink:
> 
> https://www.dslreports.com/speedtest/70320015
> https://www.dslreports.com/speedtest/70346586
> https://www.dslreports.com/speedtest/70346578

Thanks!

Best Regards
	Sebastian



> 
> /Jonas
> 
> On Fri, Jan 14, 2022 at 12:55 PM Sebastian Moeller <moeller0@gmx.de> wrote:
> Hi Jonas,
> 
> 
> > On Jan 14, 2022, at 11:44, Jonas Mårtensson <martensson.jonas@gmail.com> wrote:
> > 
> > Hi,
> > 
> > >  getting gpon more right has increasingly been on my mind
> > 
> > I think more right is to not turn the fiber into a shared medium in the first place but since gpon is so popular, improving it seems like a nice goal.
> > 
> > > Nobody in their right mind is going to hook up 128 terminalt to one OLT port, I hope...
> > 
> > Well, sharing one OLT port between many terminals is kind of the (only) advantage of PON, although split ratios of 32 or 64 are more typical. But often it's the loss budget that limits the ratio.
> 
>         Sure, but given that you probably need a few splices along the way and preferably pluggable connectors at both ends the loss budget is not that large (assuming an ISP does not want to push its luck and allows for stuff like end-users not cleaning the plug diligently before each plugging).
> 
> 
> > 
> > >  Fist question might to be "how broken is GPON/XGPON" to start with, no?
> > 
> > Looking at dslreports bufferbloat results for fiber, there are many samples with >250ms latency on the uplink. Unfortunately, this graph doesn't show results for 500Mbit/s or 1Gbit/s services but it's still interesting to look at:
> > 
> > https://www.dslreports.com/speedtest/results/bufferbloat?up=1
> 
>         Dslreports has no cue what a link is actually using, all it reports wich test profile a user selected, and some users like me ignore the names and simply use/recommend the profile with the desired number of flows. Plus quite a number of dedidedly metallic access technology are marketed with fiber somewhere in the name, potentially confusing users into selecting the "wrong" profile (think Fiber to the Cabinet for copper DSL or even Hybrid-Fiber-Coax for docsis cable)... in short the abels are nice, but I would not read too much inside those.
> 
> 
> > 
> > >  this thread https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/ (in German) has some instructions how to get root on one type of SFP ONU...
> > 
> > Thanks, that's an interesting thread. "Hacking" SFP ONUs seems like a popular hobby. Here are some other resources I found:
> > 
> > https://github.com/zry98/SFP-GPON-ONU
> > https://github.com/hwti/G-010S-A
> > https://forum.mikrotik.com/viewtopic.php?t=116364&start=300#p771961
> 
> Thanks for the links!
> 
> Regards
>         Sebastian
> 
> 
> 
> 
> 
> > 
> > > so pulling  a testbed together of some sort would be cool, and for that matter, having a SFP that could go right into a SFP enabled home router rather than a separate unit seems like a good idea, also
> > 
> > Yes, but ideally I guess you would also need some control of the OLT side. You may want to look into the VOLTHA project run by ONF:
> > 
> > https://wiki.opennetworking.org/display/COM/VOLTHA
> > 
> > /Jonas
> > 
> > On Thu, Jan 13, 2022 at 5:29 PM Sebastian Moeller <moeller0@gmx.de> wrote:
> > Hi Dave,
> > 
> > 
> > 
> > > On Jan 13, 2022, at 16:59, Dave Taht <dave.taht@gmail.com> wrote:
> > > 
> > > On Thu, Jan 13, 2022 at 7:57 AM Sebastian Moeller <moeller0@gmx.de> wrote:
> > >> 
> > >> Hi Dave,
> > >> 
> > >> 
> > >> this thread https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/ (in German) has some instructions how to get root on one type of SFP ONU... (I was monitoring that thread for general interest, turns out the intel falcon plattform seems somehow based on an ancient OpenWrt)
> > >> 
> > >> Regards
> > >>        Sebastian
> > > 
> > > It's really remarkable how many places are running an ancient openwrt.
> > > Starlink's use was not an abomination, but a persistent reality. Given
> > > how much
> > > chaos calmer I've found, I sometimes wish we'd somehow started the
> > > cerowrt project 2 years earlier.
> > 
> >         Yes and no.
> > 
> > > Then we'd be done by now.
> > 
> >         Hopefully, but then I would not have noticed the whole thing and would probably not have participated... ;)
> > 
> > Regards
> >         Sebastian 
> > 
> > 
> > 
> > > 
> > >> 
> > >>> On Jan 13, 2022, at 15:38, Dave Taht <dave.taht@gmail.com> wrote:
> > >>> 
> > >>> And a gpon onu
> > >>> 
> > >>> https://www.fs.com/products/133619.html
> > >>> 
> > >>> On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de> wrote:
> > >>>> 
> > >>>> That is similar to what happens in some GPON-ONT SFPs, some run a full small Linux distribution like OpenWrt inside.... though for ethernet that is unexpected.
> > >>>> This is also similar to SFP VDSL "modems" which likely run their own embedded OS as well inside the SFP package (at a time there was even a PCI VDSL2 "modem" that was actually running its own embedded system on the PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).
> > >>>> 
> > >>>> Regards
> > >>>>       Sebastian
> > >>>> 
> > >>>> 
> > >>>> 
> > >>>>> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
> > >>>>> 
> > >>>>> running linux, of course.
> > >>>>> 
> > >>>>> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
> > >>>>> 
> > >>>>> --
> > >>>>> I tried to build a better future, a few times:
> > >>>>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> > >>>>> 
> > >>>>> Dave Täht CEO, TekLibre, LLC
> > >>>>> _______________________________________________
> > >>>>> Cerowrt-devel mailing list
> > >>>>> Cerowrt-devel@lists.bufferbloat.net
> > >>>>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
> > >>>> 
> > >>> 
> > >>> 
> > >>> --
> > >>> I tried to build a better future, a few times:
> > >>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> > >>> 
> > >>> Dave Täht CEO, TekLibre, LLC
> > >> 
> > > 
> > > 
> > > -- 
> > > I tried to build a better future, a few times:
> > > https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> > > 
> > > Dave Täht CEO, TekLibre, LLC
> > 
> > _______________________________________________
> > Cerowrt-devel mailing list
> > Cerowrt-devel@lists.bufferbloat.net
> > https://lists.bufferbloat.net/listinfo/cerowrt-devel
> 

Re: [Cerowrt-devel] a smart SFP

[Jonas Mårtensson]

[-- Attachment #1: Type: text/plain, Size: 11256 bytes --]

>  Not all ISPs work with a single level of splitters, some do 1:4 and
split each of the four up into more later, I would guess that 1:4 + 1:8
will have a bigger aggregate attenuation than going 1:32 directly, but I am
guessing here, so thanks for the numbers.

It will be slightly higher, mainly due to the extra splices I think, but
other than that there is no fundamental difference between doing 1:32 and
doing 1:4 + 1:8. The fundamental loss is 3 dB per 1:2 split.

> In Germany the law is as it is and requires a passive handover point of
an ISPs network to the home network and also the freedom to choose routers
(and probably also ONTs/ONUs but that is still in flux), so plugs will be
common as will be exchange of ONTs/ONUs by end-users.

That's interesting. I would think that exchanging ONUs is quite tricky (as
also indicated by the "hacking" threads discussed earlier) since OLTs and
ONUs from different suppliers may not even be compatible. Can't the
"passive handover point" be between the ONU and the home router? In Sweden
(where FTTH is very common but PON is almost non-existing), the ISP
normally installs both an Ethernet "switch" terminating the fiber
(typically using an SFP) and a home router connected to the switch using
copper Ethernet cable. While it is common to exchange the home router (or
selecting to not install the ISP's router from the beginning), I don't
think anyone exchanges the switch with the fiber termination and doing so
would probably not work anyway since the switch needs to be "managed" by
the ISP.

/Jonas

On Fri, Jan 14, 2022 at 2:23 PM Sebastian Moeller <moeller0@gmx.de> wrote:

> Hi Jonas,
>
>
>
> > On Jan 14, 2022, at 14:12, Jonas Mårtensson <martensson.jonas@gmail.com>
> wrote:
> >
> > >  Sure, but given that you probably need a few splices along the way
> and preferably pluggable connectors at both ends the loss budget is not
> that large (assuming an ISP does not want to push its luck and allows for
> stuff like end-users not cleaning the plug diligently before each plugging).
> >
> > GPON loss budget is 28 dB and typical insertion losses for 1:32 and 1:64
> splitters are 17 dB and 21 dB. This leaves 11 dB or 7 dB for splices,
> connectors and fiber losses. I don't think it's common to have end-users
> cleaning and plugging in the fiber, this is done by the ISP technician at
> installation.
>
>         Not all ISPs work with a single level of splitters, some do 1:4
> and split each of the four up into more later, I would guess that 1:4 + 1:8
> will have a bigger aggregate attenuation than going 1:32 directly, but I am
> guessing here, so thanks for the numbers. In Germany the law is as it is
> and requires a passive handover point of an ISPs network to the home
> network and also the freedom to choose routers (and probably also ONTs/ONUs
> but that is still in flux), so plugs will be common as will be exchange of
> ONTs/ONUs by end-users. Whether that is a good or a bad thins is open for
> discussion, but as an ISP I would try to plan my PON plant such that there
> would be more loss reserve for these final connections than would be if
> these would be performed and documented by trained technicians. (I think
> that might be one of the consequences of deploying FTTH in the mass market,
> solutions need to be a bit more error tolerant than for networks mainly
> handled by experts only.)
>
> >
> > > Dslreports has no cue what a link is actually using, all it reports
> wich test profile a user selected, and some users like me ignore the names
> and simply use/recommend the profile with the desired number of flows. Plus
> quite a number of dedidedly metallic access technology are marketed with
> fiber somewhere in the name, potentially confusing users into selecting the
> "wrong" profile (think Fiber to the Cabinet for copper DSL or even
> Hybrid-Fiber-Coax for docsis cable)... in short the abels are nice, but I
> would not read too much inside those.
> >
> > I agree with all these points. It may be better to look at ISPs that are
> known to only use PON, such as Google Fiber. Here are some recent tests
> that all show similar and interesting bufferbloat behaviour on the uplink:
> >
> > https://www.dslreports.com/speedtest/70320015
> > https://www.dslreports.com/speedtest/70346586
> > https://www.dslreports.com/speedtest/70346578
>
> Thanks!
>
> Best Regards
>         Sebastian
>
>
>
> >
> > /Jonas
> >
> > On Fri, Jan 14, 2022 at 12:55 PM Sebastian Moeller <moeller0@gmx.de>
> wrote:
> > Hi Jonas,
> >
> >
> > > On Jan 14, 2022, at 11:44, Jonas Mårtensson <
> martensson.jonas@gmail.com> wrote:
> > >
> > > Hi,
> > >
> > > >  getting gpon more right has increasingly been on my mind
> > >
> > > I think more right is to not turn the fiber into a shared medium in
> the first place but since gpon is so popular, improving it seems like a
> nice goal.
> > >
> > > > Nobody in their right mind is going to hook up 128 terminalt to one
> OLT port, I hope...
> > >
> > > Well, sharing one OLT port between many terminals is kind of the
> (only) advantage of PON, although split ratios of 32 or 64 are more
> typical. But often it's the loss budget that limits the ratio.
> >
> >         Sure, but given that you probably need a few splices along the
> way and preferably pluggable connectors at both ends the loss budget is not
> that large (assuming an ISP does not want to push its luck and allows for
> stuff like end-users not cleaning the plug diligently before each plugging).
> >
> >
> > >
> > > >  Fist question might to be "how broken is GPON/XGPON" to start with,
> no?
> > >
> > > Looking at dslreports bufferbloat results for fiber, there are many
> samples with >250ms latency on the uplink. Unfortunately, this graph
> doesn't show results for 500Mbit/s or 1Gbit/s services but it's still
> interesting to look at:
> > >
> > > https://www.dslreports.com/speedtest/results/bufferbloat?up=1
> >
> >         Dslreports has no cue what a link is actually using, all it
> reports wich test profile a user selected, and some users like me ignore
> the names and simply use/recommend the profile with the desired number of
> flows. Plus quite a number of dedidedly metallic access technology are
> marketed with fiber somewhere in the name, potentially confusing users into
> selecting the "wrong" profile (think Fiber to the Cabinet for copper DSL or
> even Hybrid-Fiber-Coax for docsis cable)... in short the abels are nice,
> but I would not read too much inside those.
> >
> >
> > >
> > > >  this thread
> https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/
> (in German) has some instructions how to get root on one type of SFP ONU...
> > >
> > > Thanks, that's an interesting thread. "Hacking" SFP ONUs seems like a
> popular hobby. Here are some other resources I found:
> > >
> > > https://github.com/zry98/SFP-GPON-ONU
> > > https://github.com/hwti/G-010S-A
> > > https://forum.mikrotik.com/viewtopic.php?t=116364&start=300#p771961
> >
> > Thanks for the links!
> >
> > Regards
> >         Sebastian
> >
> >
> >
> >
> >
> > >
> > > > so pulling  a testbed together of some sort would be cool, and for
> that matter, having a SFP that could go right into a SFP enabled home
> router rather than a separate unit seems like a good idea, also
> > >
> > > Yes, but ideally I guess you would also need some control of the OLT
> side. You may want to look into the VOLTHA project run by ONF:
> > >
> > > https://wiki.opennetworking.org/display/COM/VOLTHA
> > >
> > > /Jonas
> > >
> > > On Thu, Jan 13, 2022 at 5:29 PM Sebastian Moeller <moeller0@gmx.de>
> wrote:
> > > Hi Dave,
> > >
> > >
> > >
> > > > On Jan 13, 2022, at 16:59, Dave Taht <dave.taht@gmail.com> wrote:
> > > >
> > > > On Thu, Jan 13, 2022 at 7:57 AM Sebastian Moeller <moeller0@gmx.de>
> wrote:
> > > >>
> > > >> Hi Dave,
> > > >>
> > > >>
> > > >> this thread
> https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/
> (in German) has some instructions how to get root on one type of SFP ONU...
> (I was monitoring that thread for general interest, turns out the intel
> falcon plattform seems somehow based on an ancient OpenWrt)
> > > >>
> > > >> Regards
> > > >>        Sebastian
> > > >
> > > > It's really remarkable how many places are running an ancient
> openwrt.
> > > > Starlink's use was not an abomination, but a persistent reality.
> Given
> > > > how much
> > > > chaos calmer I've found, I sometimes wish we'd somehow started the
> > > > cerowrt project 2 years earlier.
> > >
> > >         Yes and no.
> > >
> > > > Then we'd be done by now.
> > >
> > >         Hopefully, but then I would not have noticed the whole thing
> and would probably not have participated... ;)
> > >
> > > Regards
> > >         Sebastian
> > >
> > >
> > >
> > > >
> > > >>
> > > >>> On Jan 13, 2022, at 15:38, Dave Taht <dave.taht@gmail.com> wrote:
> > > >>>
> > > >>> And a gpon onu
> > > >>>
> > > >>> https://www.fs.com/products/133619.html
> > > >>>
> > > >>> On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de>
> wrote:
> > > >>>>
> > > >>>> That is similar to what happens in some GPON-ONT SFPs, some run a
> full small Linux distribution like OpenWrt inside.... though for ethernet
> that is unexpected.
> > > >>>> This is also similar to SFP VDSL "modems" which likely run their
> own embedded OS as well inside the SFP package (at a time there was even a
> PCI VDSL2 "modem" that was actually running its own embedded system on the
> PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).
> > > >>>>
> > > >>>> Regards
> > > >>>>       Sebastian
> > > >>>>
> > > >>>>
> > > >>>>
> > > >>>>> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com>
> wrote:
> > > >>>>>
> > > >>>>> running linux, of course.
> > > >>>>>
> > > >>>>> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
> > > >>>>>
> > > >>>>> --
> > > >>>>> I tried to build a better future, a few times:
> > > >>>>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> > > >>>>>
> > > >>>>> Dave Täht CEO, TekLibre, LLC
> > > >>>>> _______________________________________________
> > > >>>>> Cerowrt-devel mailing list
> > > >>>>> Cerowrt-devel@lists.bufferbloat.net
> > > >>>>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
> > > >>>>
> > > >>>
> > > >>>
> > > >>> --
> > > >>> I tried to build a better future, a few times:
> > > >>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> > > >>>
> > > >>> Dave Täht CEO, TekLibre, LLC
> > > >>
> > > >
> > > >
> > > > --
> > > > I tried to build a better future, a few times:
> > > > https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> > > >
> > > > Dave Täht CEO, TekLibre, LLC
> > >
> > > _______________________________________________
> > > Cerowrt-devel mailing list
> > > Cerowrt-devel@lists.bufferbloat.net
> > > https://lists.bufferbloat.net/listinfo/cerowrt-devel
> >
>
>

[-- Attachment #2: Type: text/html, Size: 16007 bytes --]

Re: [Cerowrt-devel] a smart SFP

[Sebastian Moeller]

Hi Jonas,


> On Jan 14, 2022, at 14:50, Jonas Mårtensson <martensson.jonas@gmail.com> wrote:
> 
> >  Not all ISPs work with a single level of splitters, some do 1:4 and split each of the four up into more later, I would guess that 1:4 + 1:8 will have a bigger aggregate attenuation than going 1:32 directly, but I am guessing here, so thanks for the numbers.
> 
> It will be slightly higher, mainly due to the extra splices I think, but other than that there is no fundamental difference between doing 1:32 and doing 1:4 + 1:8. The fundamental loss is 3 dB per 1:2 split.
> 
> > In Germany the law is as it is and requires a passive handover point of an ISPs network to the home network and also the freedom to choose routers (and probably also ONTs/ONUs but that is still in flux), so plugs will be common as will be exchange of ONTs/ONUs by end-users.
> 
> That's interesting. I would think that exchanging ONUs is quite tricky (as also indicated by the "hacking" threads discussed earlier) since OLTs and ONUs from different suppliers may not even be compatible.

	Well the big incumbent has a process under test which might require one phone call to transmit one important number after which a compatible ONT (there is a compatibility list I think) can and will be provisioned, so this is possible. And given that GPON is an ITU standard I would assume that unless vendoe-specific extra features are used ONTs should be compatible.


> Can't the "passive handover point" be between the ONU and the home router?

	That is under discussion. The BEREC describes three hand-over points, one passive before the ONT, one at the ethernet port of an ONT (or maybe the cage-end of an SFP-module) and/or behind an ISP supplied router (see https://berec.europa.eu/eng/document_register/subject_matter/berec/regulatory_best_practices/guidelines/9033-berec-guidelines-on-common-approaches-to-the-identification-of-the-network-termination-point-in-different-network-topologies). The German law however is clear, the termination point needs to be passive (unless there is an explicit exemption granted by the national regulatory authority NRA) OR the ISP declares an device as part of its active network, but then the ISP needs to cover the electricity cost of that device. The NRA BNetzA has communicated that it expects ONT to be treated the same as routers (for which the law grants freedom of choice by end-users) and that it is unlikely/unwilling to grant exceptions unless there are excellent reasons to do so (so far ISPs did not bring forward convincing arguments for not offering a passive termination point, as far as the BNetzA is concerned).


> In Sweden (where FTTH is very common but PON is almost non-existing), the ISP normally installs both an Ethernet "switch" terminating the fiber (typically using an SFP) and a home router connected to the switch using copper Ethernet cable. While it is common to exchange the home router (or selecting to not install the ISP's router from the beginning), I don't think anyone exchanges the switch with the fiber termination and doing so would probably not work anyway since the switch needs to be "managed" by the ISP.

	German law allows for that, as long as the ISP covers the electricity bill of that switch (operating it as part of the ISPs own active network), the termination point then would be the CatN-socket in each flat. The goal in Germany is true FTTH where each dwelling unit/flat has its own fiber connection (and then AON becomes pricy, you need to run all those fibers to COs and then have enough room for the required patching/splicing, with fiber taking more room than coper).


Regards
	Sebastian


P.S.: While AON is the technically better solution (you can always easilyput a splitter in a CO and run X direct fibers as a PON, but try the same when the spillters are out in the field much closer to the end-users than the CO and there are not enough fibers between splitter and CO). Telcos especially incumbents however prefer PON for a number of reasons:
a) price, it simply is cheaper, fewer fibers to pull and terminate and OLTs appear smaller and cheaper that active switches that can supply a similar number of end-nodes.
b) control I: if a telco builds a PON plant every user of that PON will owe the incumbent some money
c) control II: no competitor will be able to offer more advanced technology over a PON than its owner.

> 
> /Jonas
> 
> On Fri, Jan 14, 2022 at 2:23 PM Sebastian Moeller <moeller0@gmx.de> wrote:
> Hi Jonas,
> 
> 
> 
> > On Jan 14, 2022, at 14:12, Jonas Mårtensson <martensson.jonas@gmail.com> wrote:
> > 
> > >  Sure, but given that you probably need a few splices along the way and preferably pluggable connectors at both ends the loss budget is not that large (assuming an ISP does not want to push its luck and allows for stuff like end-users not cleaning the plug diligently before each plugging).
> > 
> > GPON loss budget is 28 dB and typical insertion losses for 1:32 and 1:64 splitters are 17 dB and 21 dB. This leaves 11 dB or 7 dB for splices, connectors and fiber losses. I don't think it's common to have end-users cleaning and plugging in the fiber, this is done by the ISP technician at installation.
> 
>         Not all ISPs work with a single level of splitters, some do 1:4 and split each of the four up into more later, I would guess that 1:4 + 1:8 will have a bigger aggregate attenuation than going 1:32 directly, but I am guessing here, so thanks for the numbers. In Germany the law is as it is and requires a passive handover point of an ISPs network to the home network and also the freedom to choose routers (and probably also ONTs/ONUs but that is still in flux), so plugs will be common as will be exchange of ONTs/ONUs by end-users. Whether that is a good or a bad thins is open for discussion, but as an ISP I would try to plan my PON plant such that there would be more loss reserve for these final connections than would be if these would be performed and documented by trained technicians. (I think that might be one of the consequences of deploying FTTH in the mass market, solutions need to be a bit more error tolerant than for networks mainly handled by experts only.)
> 
> > 
> > > Dslreports has no cue what a link is actually using, all it reports wich test profile a user selected, and some users like me ignore the names and simply use/recommend the profile with the desired number of flows. Plus quite a number of dedidedly metallic access technology are marketed with fiber somewhere in the name, potentially confusing users into selecting the "wrong" profile (think Fiber to the Cabinet for copper DSL or even Hybrid-Fiber-Coax for docsis cable)... in short the abels are nice, but I would not read too much inside those.
> > 
> > I agree with all these points. It may be better to look at ISPs that are known to only use PON, such as Google Fiber. Here are some recent tests that all show similar and interesting bufferbloat behaviour on the uplink:
> > 
> > https://www.dslreports.com/speedtest/70320015
> > https://www.dslreports.com/speedtest/70346586
> > https://www.dslreports.com/speedtest/70346578
> 
> Thanks!
> 
> Best Regards
>         Sebastian
> 
> 
> 
> > 
> > /Jonas
> > 
> > On Fri, Jan 14, 2022 at 12:55 PM Sebastian Moeller <moeller0@gmx.de> wrote:
> > Hi Jonas,
> > 
> > 
> > > On Jan 14, 2022, at 11:44, Jonas Mårtensson <martensson.jonas@gmail.com> wrote:
> > > 
> > > Hi,
> > > 
> > > >  getting gpon more right has increasingly been on my mind
> > > 
> > > I think more right is to not turn the fiber into a shared medium in the first place but since gpon is so popular, improving it seems like a nice goal.
> > > 
> > > > Nobody in their right mind is going to hook up 128 terminalt to one OLT port, I hope...
> > > 
> > > Well, sharing one OLT port between many terminals is kind of the (only) advantage of PON, although split ratios of 32 or 64 are more typical. But often it's the loss budget that limits the ratio.
> > 
> >         Sure, but given that you probably need a few splices along the way and preferably pluggable connectors at both ends the loss budget is not that large (assuming an ISP does not want to push its luck and allows for stuff like end-users not cleaning the plug diligently before each plugging).
> > 
> > 
> > > 
> > > >  Fist question might to be "how broken is GPON/XGPON" to start with, no?
> > > 
> > > Looking at dslreports bufferbloat results for fiber, there are many samples with >250ms latency on the uplink. Unfortunately, this graph doesn't show results for 500Mbit/s or 1Gbit/s services but it's still interesting to look at:
> > > 
> > > https://www.dslreports.com/speedtest/results/bufferbloat?up=1
> > 
> >         Dslreports has no cue what a link is actually using, all it reports wich test profile a user selected, and some users like me ignore the names and simply use/recommend the profile with the desired number of flows. Plus quite a number of dedidedly metallic access technology are marketed with fiber somewhere in the name, potentially confusing users into selecting the "wrong" profile (think Fiber to the Cabinet for copper DSL or even Hybrid-Fiber-Coax for docsis cable)... in short the abels are nice, but I would not read too much inside those.
> > 
> > 
> > > 
> > > >  this thread https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/ (in German) has some instructions how to get root on one type of SFP ONU...
> > > 
> > > Thanks, that's an interesting thread. "Hacking" SFP ONUs seems like a popular hobby. Here are some other resources I found:
> > > 
> > > https://github.com/zry98/SFP-GPON-ONU
> > > https://github.com/hwti/G-010S-A
> > > https://forum.mikrotik.com/viewtopic.php?t=116364&start=300#p771961
> > 
> > Thanks for the links!
> > 
> > Regards
> >         Sebastian
> > 
> > 
> > 
> > 
> > 
> > > 
> > > > so pulling  a testbed together of some sort would be cool, and for that matter, having a SFP that could go right into a SFP enabled home router rather than a separate unit seems like a good idea, also
> > > 
> > > Yes, but ideally I guess you would also need some control of the OLT side. You may want to look into the VOLTHA project run by ONF:
> > > 
> > > https://wiki.opennetworking.org/display/COM/VOLTHA
> > > 
> > > /Jonas
> > > 
> > > On Thu, Jan 13, 2022 at 5:29 PM Sebastian Moeller <moeller0@gmx.de> wrote:
> > > Hi Dave,
> > > 
> > > 
> > > 
> > > > On Jan 13, 2022, at 16:59, Dave Taht <dave.taht@gmail.com> wrote:
> > > > 
> > > > On Thu, Jan 13, 2022 at 7:57 AM Sebastian Moeller <moeller0@gmx.de> wrote:
> > > >> 
> > > >> Hi Dave,
> > > >> 
> > > >> 
> > > >> this thread https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/ (in German) has some instructions how to get root on one type of SFP ONU... (I was monitoring that thread for general interest, turns out the intel falcon plattform seems somehow based on an ancient OpenWrt)
> > > >> 
> > > >> Regards
> > > >>        Sebastian
> > > > 
> > > > It's really remarkable how many places are running an ancient openwrt.
> > > > Starlink's use was not an abomination, but a persistent reality. Given
> > > > how much
> > > > chaos calmer I've found, I sometimes wish we'd somehow started the
> > > > cerowrt project 2 years earlier.
> > > 
> > >         Yes and no.
> > > 
> > > > Then we'd be done by now.
> > > 
> > >         Hopefully, but then I would not have noticed the whole thing and would probably not have participated... ;)
> > > 
> > > Regards
> > >         Sebastian 
> > > 
> > > 
> > > 
> > > > 
> > > >> 
> > > >>> On Jan 13, 2022, at 15:38, Dave Taht <dave.taht@gmail.com> wrote:
> > > >>> 
> > > >>> And a gpon onu
> > > >>> 
> > > >>> https://www.fs.com/products/133619.html
> > > >>> 
> > > >>> On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de> wrote:
> > > >>>> 
> > > >>>> That is similar to what happens in some GPON-ONT SFPs, some run a full small Linux distribution like OpenWrt inside.... though for ethernet that is unexpected.
> > > >>>> This is also similar to SFP VDSL "modems" which likely run their own embedded OS as well inside the SFP package (at a time there was even a PCI VDSL2 "modem" that was actually running its own embedded system on the PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).
> > > >>>> 
> > > >>>> Regards
> > > >>>>       Sebastian
> > > >>>> 
> > > >>>> 
> > > >>>> 
> > > >>>>> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
> > > >>>>> 
> > > >>>>> running linux, of course.
> > > >>>>> 
> > > >>>>> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
> > > >>>>> 
> > > >>>>> --
> > > >>>>> I tried to build a better future, a few times:
> > > >>>>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> > > >>>>> 
> > > >>>>> Dave Täht CEO, TekLibre, LLC
> > > >>>>> _______________________________________________
> > > >>>>> Cerowrt-devel mailing list
> > > >>>>> Cerowrt-devel@lists.bufferbloat.net
> > > >>>>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
> > > >>>> 
> > > >>> 
> > > >>> 
> > > >>> --
> > > >>> I tried to build a better future, a few times:
> > > >>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> > > >>> 
> > > >>> Dave Täht CEO, TekLibre, LLC
> > > >> 
> > > > 
> > > > 
> > > > -- 
> > > > I tried to build a better future, a few times:
> > > > https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
> > > > 
> > > > Dave Täht CEO, TekLibre, LLC
> > > 
> > > _______________________________________________
> > > Cerowrt-devel mailing list
> > > Cerowrt-devel@lists.bufferbloat.net
> > > https://lists.bufferbloat.net/listinfo/cerowrt-devel
> > 
> 

Re: [Cerowrt-devel] a smart SFP

[Dave Taht]

On Fri, Jan 14, 2022 at 5:12 AM Jonas Mårtensson
<martensson.jonas@gmail.com> wrote:
>
> >  Sure, but given that you probably need a few splices along the way and preferably pluggable connectors at both ends the loss budget is not that large (assuming an ISP does not want to push its luck and allows for stuff like end-users not cleaning the plug diligently before each plugging).
>
> GPON loss budget is 28 dB and typical insertion losses for 1:32 and 1:64 splitters are 17 dB and 21 dB. This leaves 11 dB or 7 dB for splices, connectors and fiber losses. I don't think it's common to have end-users cleaning and plugging in the fiber, this is done by the ISP technician at installation.
>
> > Dslreports has no cue what a link is actually using, all it reports wich test profile a user selected, and some users like me ignore the names and simply use/recommend the profile with the desired number of flows. Plus quite a number of dedidedly metallic access technology are marketed with fiber somewhere in the name, potentially confusing users into selecting the "wrong" profile (think Fiber to the Cabinet for copper DSL or even Hybrid-Fiber-Coax for docsis cable)... in short the abels are nice, but I would not read too much inside those.
>
> I agree with all these points. It may be better to look at ISPs that are known to only use PON, such as Google Fiber. Here are some recent tests that all show similar and interesting bufferbloat behaviour on the uplink:
>
> https://www.dslreports.com/speedtest/70320015

This is gfiber. We didn't so far as I remember get fq_codel going on
the uplink there (just the wifi)

I don't know anybody anymore over there.

> https://www.dslreports.com/speedtest/70346586
> https://www.dslreports.com/speedtest/70346578
>
> /Jonas
>
> On Fri, Jan 14, 2022 at 12:55 PM Sebastian Moeller <moeller0@gmx.de> wrote:
>>
>> Hi Jonas,
>>
>>
>> > On Jan 14, 2022, at 11:44, Jonas Mårtensson <martensson.jonas@gmail.com> wrote:
>> >
>> > Hi,
>> >
>> > >  getting gpon more right has increasingly been on my mind
>> >
>> > I think more right is to not turn the fiber into a shared medium in the first place but since gpon is so popular, improving it seems like a nice goal.
>> >
>> > > Nobody in their right mind is going to hook up 128 terminalt to one OLT port, I hope...
>> >
>> > Well, sharing one OLT port between many terminals is kind of the (only) advantage of PON, although split ratios of 32 or 64 are more typical. But often it's the loss budget that limits the ratio.
>>
>>         Sure, but given that you probably need a few splices along the way and preferably pluggable connectors at both ends the loss budget is not that large (assuming an ISP does not want to push its luck and allows for stuff like end-users not cleaning the plug diligently before each plugging).
>>
>>
>> >
>> > >  Fist question might to be "how broken is GPON/XGPON" to start with, no?
>> >
>> > Looking at dslreports bufferbloat results for fiber, there are many samples with >250ms latency on the uplink. Unfortunately, this graph doesn't show results for 500Mbit/s or 1Gbit/s services but it's still interesting to look at:
>> >
>> > https://www.dslreports.com/speedtest/results/bufferbloat?up=1
>>
>>         Dslreports has no cue what a link is actually using, all it reports wich test profile a user selected, and some users like me ignore the names and simply use/recommend the profile with the desired number of flows. Plus quite a number of dedidedly metallic access technology are marketed with fiber somewhere in the name, potentially confusing users into selecting the "wrong" profile (think Fiber to the Cabinet for copper DSL or even Hybrid-Fiber-Coax for docsis cable)... in short the abels are nice, but I would not read too much inside those.
>>
>>
>> >
>> > >  this thread https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/ (in German) has some instructions how to get root on one type of SFP ONU...
>> >
>> > Thanks, that's an interesting thread. "Hacking" SFP ONUs seems like a popular hobby. Here are some other resources I found:
>> >
>> > https://github.com/zry98/SFP-GPON-ONU
>> > https://github.com/hwti/G-010S-A
>> > https://forum.mikrotik.com/viewtopic.php?t=116364&start=300#p771961
>>
>> Thanks for the links!
>>
>> Regards
>>         Sebastian
>>
>>
>>
>>
>>
>> >
>> > > so pulling  a testbed together of some sort would be cool, and for that matter, having a SFP that could go right into a SFP enabled home router rather than a separate unit seems like a good idea, also
>> >
>> > Yes, but ideally I guess you would also need some control of the OLT side. You may want to look into the VOLTHA project run by ONF:
>> >
>> > https://wiki.opennetworking.org/display/COM/VOLTHA
>> >
>> > /Jonas
>> >
>> > On Thu, Jan 13, 2022 at 5:29 PM Sebastian Moeller <moeller0@gmx.de> wrote:
>> > Hi Dave,
>> >
>> >
>> >
>> > > On Jan 13, 2022, at 16:59, Dave Taht <dave.taht@gmail.com> wrote:
>> > >
>> > > On Thu, Jan 13, 2022 at 7:57 AM Sebastian Moeller <moeller0@gmx.de> wrote:
>> > >>
>> > >> Hi Dave,
>> > >>
>> > >>
>> > >> this thread https://www.computerbase.de/forum/threads/eigenes-modem-an-ftth-anschluss-via-sfp-gpon-modul.2061989/ (in German) has some instructions how to get root on one type of SFP ONU... (I was monitoring that thread for general interest, turns out the intel falcon plattform seems somehow based on an ancient OpenWrt)
>> > >>
>> > >> Regards
>> > >>        Sebastian
>> > >
>> > > It's really remarkable how many places are running an ancient openwrt.
>> > > Starlink's use was not an abomination, but a persistent reality. Given
>> > > how much
>> > > chaos calmer I've found, I sometimes wish we'd somehow started the
>> > > cerowrt project 2 years earlier.
>> >
>> >         Yes and no.
>> >
>> > > Then we'd be done by now.
>> >
>> >         Hopefully, but then I would not have noticed the whole thing and would probably not have participated... ;)
>> >
>> > Regards
>> >         Sebastian
>> >
>> >
>> >
>> > >
>> > >>
>> > >>> On Jan 13, 2022, at 15:38, Dave Taht <dave.taht@gmail.com> wrote:
>> > >>>
>> > >>> And a gpon onu
>> > >>>
>> > >>> https://www.fs.com/products/133619.html
>> > >>>
>> > >>> On Thu, Jan 13, 2022 at 6:23 AM Sebastian Moeller <moeller0@gmx.de> wrote:
>> > >>>>
>> > >>>> That is similar to what happens in some GPON-ONT SFPs, some run a full small Linux distribution like OpenWrt inside.... though for ethernet that is unexpected.
>> > >>>> This is also similar to SFP VDSL "modems" which likely run their own embedded OS as well inside the SFP package (at a time there was even a PCI VDSL2 "modem" that was actually running its own embedded system on the PCI board, IIRC, it pretended to the main computer to be an ethernet NIC).
>> > >>>>
>> > >>>> Regards
>> > >>>>       Sebastian
>> > >>>>
>> > >>>>
>> > >>>>
>> > >>>>> On Jan 13, 2022, at 15:18, Dave Taht <dave.taht@gmail.com> wrote:
>> > >>>>>
>> > >>>>> running linux, of course.
>> > >>>>>
>> > >>>>> https://blog.benjojo.co.uk/post/smart-sfp-linux-inside
>> > >>>>>
>> > >>>>> --
>> > >>>>> I tried to build a better future, a few times:
>> > >>>>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
>> > >>>>>
>> > >>>>> Dave Täht CEO, TekLibre, LLC
>> > >>>>> _______________________________________________
>> > >>>>> Cerowrt-devel mailing list
>> > >>>>> Cerowrt-devel@lists.bufferbloat.net
>> > >>>>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>> > >>>>
>> > >>>
>> > >>>
>> > >>> --
>> > >>> I tried to build a better future, a few times:
>> > >>> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
>> > >>>
>> > >>> Dave Täht CEO, TekLibre, LLC
>> > >>
>> > >
>> > >
>> > > --
>> > > I tried to build a better future, a few times:
>> > > https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
>> > >
>> > > Dave Täht CEO, TekLibre, LLC
>> >
>> > _______________________________________________
>> > Cerowrt-devel mailing list
>> > Cerowrt-devel@lists.bufferbloat.net
>> > https://lists.bufferbloat.net/listinfo/cerowrt-devel
>>


-- 
I tried to build a better future, a few times:
https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org

Dave Täht CEO, TekLibre, LLC

Re: [Cerowrt-devel] a smart SFP

[Michael Richardson]

Jonas Mårtensson <martensson.jonas@gmail.com> wrote:
    >> getting gpon more right has increasingly been on my mind

    > I think more right is to not turn the fiber into a shared medium in the
    > first place but since gpon is so popular, improving it seems like a
    > nice goal.

    >> Nobody in their right mind is going to hook up 128 terminalt to one
    >> OLT
    > port, I hope...

    > Well, sharing one OLT port between many terminals is kind of the (only)
    > advantage of PON, although split ratios of 32 or 64 are more
    > typical. But often it's the loss budget that limits the ratio.

And, there are also situations, many of them industrial, where the fiber
allows one to avoid ground loops, and where the bandwidth/latency
requirements are very modest.
it's not all youtube downloads and zoom meetings :-)

I was involved in a GPON build back in 2010 into a very poorly served
industrial park.  (That was for email/web browsing by the industries).
A reason they were so poorly served is that there were two expressways (one
of them the TransCanada) and a river that bordered the area.
There were only eight strands in a single conduit available....
We considered moving (an) OLT into that area actually... DWDM to the rescue
in the end, but that was a bit bleeding edge at the time.

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

Re: [Cerowrt-devel] a smart SFP

[Joel Wirāmu Pauling]

[-- Attachment #1: Type: text/plain, Size: 3300 bytes --]

Those are really high split rates. We (as in UFB in NZ) looked at 32:1
splits but it's rare - in practice it's often half that. Splits end up
based on contention of regulated L2 plans which are sold to RSP's to
on-sell to customers. Based on available backhaul bandwidth rather than any
factor. i.e on a 2.5Gbit Duplex GPON port which carries a L2 service to
RSP(L3 providers) it's a simple matter of dividing up the Active Optical
backhaul to the GPON unit (minimally 2x10Gbit Active Paths to the CO for an
old style GPON only node) so that gives you around 66 customers on a
300/300 service without resorting to Teletraffic engineering in the access
equipment. Splits per line card port are then distributed based on that
metric rather than anything else. There are of course some exceptions to
this rule for some nodes in the network but it's uncommon.

RSP's are responsible once the L2 is handed to them, so experience varies
once it's off the access network. But the entire PON network is engineered
without contention at any points, including in the last-mile, it actually
works out cheaper than introducing engineering required for massive
contention ratios and results in better experience for everyone.

The active GPON nodes used in the network mostly have been moved to 2 or 4
* 40gbit or 100Gbit uplinks at the this point to support XPON.


On Sat, Jan 15, 2022 at 8:22 AM Michael Richardson <mcr@sandelman.ca> wrote:

>
> Jonas Mårtensson <martensson.jonas@gmail.com> wrote:
>     >> getting gpon more right has increasingly been on my mind
>
>     > I think more right is to not turn the fiber into a shared medium in
> the
>     > first place but since gpon is so popular, improving it seems like a
>     > nice goal.
>
>     >> Nobody in their right mind is going to hook up 128 terminalt to one
>     >> OLT
>     > port, I hope...
>
>     > Well, sharing one OLT port between many terminals is kind of the
> (only)
>     > advantage of PON, although split ratios of 32 or 64 are more
>     > typical. But often it's the loss budget that limits the ratio.
>
> And, there are also situations, many of them industrial, where the fiber
> allows one to avoid ground loops, and where the bandwidth/latency
> requirements are very modest.
> it's not all youtube downloads and zoom meetings :-)
>
> I was involved in a GPON build back in 2010 into a very poorly served
> industrial park.  (That was for email/web browsing by the industries).
> A reason they were so poorly served is that there were two expressways (one
> of them the TransCanada) and a river that bordered the area.
> There were only eight strands in a single conduit available....
> We considered moving (an) OLT into that area actually... DWDM to the rescue
> in the end, but that was a bit bleeding edge at the time.
>
> --
> ]               Never tell me the odds!                 | ipv6 mesh
> networks [
> ]   Michael Richardson, Sandelman Software Works        |    IoT
> 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
>

[-- Attachment #2: Type: text/html, Size: 4571 bytes --]