Electric utilities, IP addressing, and BPL (was Re: Google wants to be your Internet)

   * From: Sean Donelan
   * Date: Tue Jan 23 15:06:02 2007

What do you do when the electric companies split up again,
renumber the meters into different network blocks?

Thanks for the discussion. It's rare I've seen a thread on NANOG
that's so pertinent to my own situation.

I work for a fairly large utility faced with these very problems. We
are developing a ~half $billion project to replace our electric meters
with 2-way communicating ones, and have several other things going,
such as "intelligent grid" projects that will grow the size of our
networks many times over. I've also been on a broadband over power
line (BPL) project for a year and a half, and have a good
understanding of how & where it works, and where it doesn't.

The US electric distribution grid, is, for the most part, quite dumb.
There are various SCADA and fault protection systems out there, but
not everything is automated, and certainly not everything is covered.
For example, when the power goes out somewhere, almost all of the
time, we don't know until we receive calls from customers. Not a
proactive system. Not the level of service that our customers want.

Intelligent communicating meters are at the core of the modernization
of the grid (and of utility profits, if you look at it in certain
ways). You'll eventually find these meters tied to time-of-use
electric rates (more $$$ during peak times), as well as used as part
of an overall instrumentation effort of the grid.

Demand response is also a piece of that puzzle. Shaving a few
megawatts on a hot day by reducing load (think 50,000 thermostats
having their temperature settings moved up a few degrees on a really
hot day) can mean avoiding building power plant and/or transmission
line capacity. The electric meter is potentially the utility's gateway
into a neighborhood or home network, and could enable those

Instrumentation of the grid will also take sensor nets, on the order
of tens or hundreds of thousands of endpoints per distribution system.
We'll find similar networks in customer homes - the load control
boxes, the thermostats, etc., that will extend from our edge.
Potentially, hundreds of thousands more networks there.

There are some utilities that will use BPL as a partial or whole
transport for metering & other utility applications, with the notion
that there's enough bandwidth to handle residential access services as
well (this will often be via a tenant or other 3rd party). While BPL
business cases won't be made without utility applications like meter
reading, it's unlikely they'll be made without an access play. The
other stuff just doesn't require the bandwidth.

Do we need IP all the way to the edge of the network? We don't think
so. But the alternatives, such as adding ZigBee mesh networks beyond
the IP edge, have their own problems of scale & complexity.

I've been following the notion that however far our IP network extends
out, IPV4 and RFC1918 have plenty of space for us. I'm reminded by
this thread, however, that M&A activities, among others, might one day
bite us for making that assumption.


Hi Josh,

According to our own experiments and real deployment experience in this
field for many years, IPv6 is the best solution. Because the lack of support
of IPv6 in many applications, we also keep IPv4+NAT, so actually is a dual
stack situation, but all new sensors, meters, and apps. in general, can
easily support IPv6 and make your life much easier.

You may not need now IP up to the edge, but sooner or later, it will change,
and also the high number of devices and networks that you want to put
together. Having enough addresses provides you end to end connectivity,
which again, may not be seen now as a real need, but it will become sooner
or later.

You can find lots of documents, from a project that we finish a couple of
years ago at http://www.6power.org. They aren't updated up to the latest
technology on this field, but sure are a good reading source.