Re: Comcast outages continue even in areas with PG&E power restored

[Brandon Martin <lists.nanog () monmotha net>]

On 10/14/19 6:38 PM, Michael Thomas wrote:
> >
> So it turns out that our local telco/isp does keep dsl running via the 
> same mechanism as they keep pots power backed up (i'm guessing it's a 
> diesel generator at the co, but am not sure). It seems that a lot of the 
> pedestals terminating the local loop these days do the conversion to IP 
> right there with sip/h.248/mgcp/rtp. I'm not sure how they get power to 
> the pedestal, but these were all a home run to the co at one time so it 
> probably wasn't hard to power them from the co. For all i know, that's 
> how they're all powered all the time, with a transfer switch at the co, 
> rather than tapping the local grid next to the pedestal.
>
> Of course this is a lot of conjecture on my part... be glad to be clued 
> in by folks in know.

Legacy carriers with existing copper plant being overbuilt by fiber 
often use "span power" at 48V or ~190VDC from the nearest central 
facility (which may be an old-school CO or RTU with utility power) to 
power their distributed plant that's deep into the network near the 
customer edge.  There's usually then also some local battery as well but 
not much.  The assumption is that the legacy copper, now being used 
simply for power, follows essentially the same routes as the fiber and 
will drop at the same time.  In a cable cut situation this is often 
true, but in a "cable rot" situation it's obviously not, and fiber paths 
aren't always the same as power paths.

The CO and older RTUs from the POTS and early ADSL era will have utility 
power as primary with somewhat extensive battery facilities.  Old-school 
CO will have a lot more battery capacity than an RTU.  An RTU will 
usually end up with a portable genset being delivered during an extended 
power outage as mentioned.  These RTUs still host legacy POTS and TDM 
services that either have serious SLAs on them or regulatory uptime 
considerations whereas distributed peds are normally best-effort, 
non-regulated data services only in many cases precisely to keep the 
reliability requirements (and therefore cost) on them down.

SBC/AT&T's pedestals they built for Lightspeed (U-Verse) do usually have 
local utility with a few hours of backup.  I'm not sure why the went 
that route rather than span power.  Might be that the early VDSL DSLAMs 
just used too much power for that.  They seem to only have maybe a day 
or so of battery before they need a portable generator brought around 
which AT&T at least has procedures for (whether they are executed or not 
is another matter).  They're something of a hybrid between a 
conventional RTU and modern distributed pedestal.

All the conventional telcos are far more focused on keeping voice 
service alive since they get raked over the coals by the FCC if it drops 
due to lack of 911.  That includes wireless if they are both a wireline 
and wireless operator.  Interestingly, VoIP service delivered to the 
customer as such (even if there's an ATA built into the customer's 
"modem") often gets a pass on this since it's not considered POTS.  Same 
goes for SLA'd business services, too especially T1s since those may 
host regulated voice.


Coax operators have historically had less need for reliability as they 
were originally built purely for convenience services (cable television) 
and have been pressed into service for more modern data needs.  Those 
"Alpha boxes" you may see around providing line power do usually have 
some batteries in them, though they're often ill-maintained and only 
provide maybe an hour of hold-up at most.

Exclusively residential areas sometimes have zero hold-up ability at 
all.  They'll drop at least outside node based digital services (e.g. 
DOCSIS) as soon as power falls along with anything being distributed 
into the field on AM fiber carriers.  It's not unheard of for 
conventional linear TV to still be delivered into the field from a 
head-end at baseband on coax and sometimes that'll stay up longer 
depending on RF power/split budget as long as the local RF head end 
still has power.


Pure-play fiber carriers, especially PON-based, get to turn what is 
often a curse in terms of design into a blessing, here.  They usually 
have almost no active outside plant or, if they do, it's less 
distributed and can afford reasonable backup power infrastructure. 
While it's annoying to have no power available as you approach customer 
prem, that forces you to make choices such that there's no worry about 
backup until you can (usually) just make it the customer's 
responsibility right at the demarc.  Your CO/head-end needs backup, of 
course, but that's usually a facility that can afford it.
--
Brandon Martin