[160325] in North American Network Operators' Group
Re: Muni fiber: L1 or L2?
daemon@ATHENA.MIT.EDU (Owen DeLong)
Mon Feb 4 00:34:22 2013
From: Owen DeLong <owen@delong.com>
In-Reply-To: <CAMrdfRxMeVXD0r70+YRO7A5Z2vuPbt9nSAygChA2MnkN2a-sHA@mail.gmail.com>
Date: Sun, 3 Feb 2013 21:30:20 -0800
To: Scott Helms <khelms@zcorum.com>
Cc: NANOG <nanog@nanog.org>
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
On Feb 3, 2013, at 2:57 PM, Scott Helms <khelms@zcorum.com> wrote:
>=20
>=20
>=20
> On Sun, Feb 3, 2013 at 5:24 PM, Owen DeLong <owen@delong.com> wrote:
>=20
> On Feb 3, 2013, at 12:33 PM, Scott Helms <khelms@zcorum.com> wrote:
>=20
>>=20
>>=20
>>=20
>> On Sun, Feb 3, 2013 at 2:53 PM, Owen DeLong <owen@delong.com> wrote:
>>=20
>> On Feb 2, 2013, at 5:06 PM, Scott Helms <khelms@zcorum.com> wrote:
>>=20
>>> Owen,
>>> I think the confusion I have is that you seem to want to create =
solutions for problems that have already been solved. There is no cost =
effective method of sharing a network at layer 1 since DWDM is expensive =
and requires compatible gear on both sides and no one has enough fiber =
(nor is cheap enough in brand new builds) to simply home run every home =
and maintain that. ISPs that would want to use the shared network in =
general (>95% in my experience) don't want to maintain the access gear =
and since there is no clear way to delineate responsibilities when there =
is an issue its hard.
>>>=20
>>>=20
>> ??
>>=20
>> Who said anything about sharing the network at L1?
>>=20
>> You did.=20
>=20
> No, I didn't. I said build out an L1 infrastructure such that =
individual connections can be leased from it. Not shared L1 connections.
> I have never advocated shared L1 connections.
>=20
> Sharing the entire network at layer 1 is what I and I believe you were =
talking about. Not sharing individual fiber connections, but using the =
same fiber plant for multiple layer 2 technologies. This is what you're =
suggesting, correct?
> =20
So long as you recognize that it's on a pair-by-pair basis end-to-end =
and not expecting any mixing/sharing/etc. by the L1 infrastructure =
provider, yes.
>=20
>>=20
>> Is it more expensive to home-run every home than to put splitters in =
the neighborhood? Yes. Is it enough more expensive that the tradeoffs =
cannot be overcome? I remain unconvinced.
>>=20
>>=20
>> This completely depends on the area and the goals of the network. In =
most cases for muni networks back hauling everything is more expensive.
>=20
> I agree it's more expensive. The question is whether it's enough more =
expensive to make it infeasible. You still haven't come anywhere near =
addressing that question.
>=20
> I've said repeatedly that this a network by network analysis. I've =
never said its infeasible, but that it is more expensive both initially =
and long term in MOST installs. That by itself is generally enough to =
invalidate the design since in almost all cases there's no benefit to =
home running all the connections. It doesn't make the connection faster =
nor do ISPs (as a group) care about a layer 1 versus layer 2 hand off.
That's where we disagree. The benefit is that:
1. It doesn't lock the entire area into a single current =
technology.
2. It allows for individual subscribers (probably mostly =
businesses, but I have had a few occasions
where this would have been useful as an individual) to =
get dark XC to other locations.
3. Subscribers who want individualized services from =
different vendors have a choice.
4. Providers have to compete on a leveled playing field and =
there is thus incentive to innovate
even if the innovation moves away from PON.
> =20
>=20
>> =20
>>=20
>> I'm not sure why you think it would be hard to delineate the =
responsibilities=85 You've got a fiber path maintained by the =
municipality with active equipment maintained by the ISP at each end. If =
the light coming out of the equipment at one end doesn't come out of the =
fiber at the other end, you have a problem in the municipality's domain. =
If the light makes it through in tact, you have a problem in the ISP's =
domain.
>>=20
>> There is equipment available that can test that fairly easily.
>>=20
>> OK, this one made my wife get scared I laughed so hard. You clearly =
have never tried to do this or had to work with different operators in =
the same physical network. Please, go talk to someone whose worked in =
the field of a FTTx network and describe this scenario to them. Its =
clear you don't want to hear it from me via email so please go do some =
research.
>=20
>=20
> I've talked to a few people doing exactly that. Yes, you need =
different test sets depending on which L2 gear is involved, but, in =
virtually ever case, there is a piece of test gear that can be used to =
test a loop independent of the configuration of the L2 gear in question.
>=20
> Yes, there is a meter for all the different kinds of technologies that =
you might want to support. For example a DOCSIS 3.0 DSAM from JDSU will =
run you around $8000.00 A PON meter with long range lasers (more than =
10 miles) from JDSU or Trilithic will cost you nearly $10,000. Exactly =
how many of those kinds of meters do you want to have to buy? How many =
of your staff are you going to train on them (they do require training =
and knowledge to use)?=20
For my proposed methods of build-out, no need for the long range lasers. =
As I said, everything should be within 8km of the MMR.
As I suggested, the simpler approach is to require the complaining L2 =
provider to cooperate in the diagnostic process and provide access to =
the applicable meters if necessary. The standard offered absent =
assistance from the L2 provider is OTDR success.
>=20
> For providers getting L1 service, it wouldn't be too hard to make this =
testing / providing necessary test equipment part of their contract.
>>> The long and short of it is lots of people have tried to L1 sharing =
and its not economical and nothing I've seen here or elsewhere changes =
that. The thing you have to remember is that muni networks have to be =
cost effective and that's not just the capital costs. The operational =
cost in the long term is much greater than the cost of initial gear and =
fiber install.
>>>=20
>> We can agree to disagree. A muni network needs to be able to recover =
its costs. The costs of building out and maintaining home-run
>> fiber are not necessarily that much greater than the costs of =
building out and maintaining fiber at the neighborhood. One option, for
>> example, would be to have neighborhood B-Boxes where the fiber can =
either be fed into provider-specific splitters (same economy
>> as existing PON deployments) or cross-connected to fiber on the F1 =
cable going back to the MMR (home-run).
>>=20
>>=20
>> We can agree all we want, that doesn't change history. Handing out =
connections at layer 1 is both more expensive and less efficient. Its =
also extremely wasteful (which is why its more expensive) since your =
lowest unit you can sell is a fiber strand whether the end customer =
wants a 3 mbps connection or a gig its the same to the city. I'm not =
saying you shouldn't sell dark fiber, I'm saying that in 99% of the =
cities you can't build a business model around doing just that unless =
your city doesn't want to break even on the build and maintenance.
>=20
> If it's $700 per home passed to build out home-run fibers (which seems =
to be a reasonable approximation from earlier discussions), then there's =
no reason you can't sell $40/month service over that where the L1 =
component is a $10/month ($7 for capital recovery, $3 for operations and =
support) pricing component.
>=20
> Nope, no reason at all if you don't care about covering your costs.=20
I just explained where the expected costs get covered, so you're going =
to have to explain that statement.
>=20
> By my estimates, to become truly impractical, you'd have to get =
somewhere north of $1500 per home passed.
>=20
>=20
>>=20
>> =20
>> The only additional cost in this system over traditional PON is the =
larger number of fibers required in the F1 cable.
>>=20
>>=20
>> PON networks aren't deployed this way and if you're going to backhaul =
all of the connections to a central point you wouldn't run PON. PON is =
worse in every performance related way to PON and the only reasons =
operators deploy it today is because its less expensive. Its less =
expensive because you don't have to backhaul all of the connections or =
have active components at the neighborhood level.
>=20
> Then don't deploy PON. I don't care whether PON gets deployed or not. =
You keep coming back at this as if PON is somehow the goal. Personally, =
I'd rather see Gig-E in every home and be done with it.=20
>=20
> However, the point is that building the infrastructure in that manner =
doesn't cost much more than building out traditional PON infrastructure =
(if you're doing it from greenfield) and it can support either =
technology.
>=20
> Sure it does, even in greenfield and whats more it costs more over the =
long term UNLESS you know where every home and business will be located =
10 years from now. =20
More yes, much more, I'm not so convinced.
Owen
