[IP] Re: how does one define "capacity"

[David Farber <dave@farber.net>]

________________________________________
From: Bob Frankston [Bob19-0501@bobf.frankston.com]
Sent: Saturday, April 19, 2008 5:39 PM
To: 'Craig Partridge'; David Farber
Subject: RE: how does one define "capacity"

This is the start of an interesting conversation …

We can argue about the limits from physics but I tend to view those as hypothetical until we get close.

DSL is a more pragmatic example. ADSL is a twenty year old technology. The capacity quadruples if you halve the length. Ciofi at Stanford argues that he can get many times the capacity within the current constraints of DSL. Of course we needn’t limit ourselves to those constraints – crosstalk can be your friend.

Information is a complex concept. What is the measure we are using for capacity? Are bits the correct measure or is the measure something as vague as "entertainment value".  That was the point of http://www.frankston.com/?name=NNSpeed. The transport doesn’t exist in isolation – it’s value comes from how we use it. That’s the social interpretation of Robert Laughlin’s A Different Universe. Scale and context matter a lot. I keep pointing out that the Internet Dynamic (http://www.frankston.com/?name=InternetDynamic) is about discovering what we can do with what is available. Thus asking the capacity is, in a sense, asking what the capacity of our imagination is.

Of course all of these assumptions reach the level of tragedy when we let regulators take them as givens and they confuse price with cost. The lawyers should know better – after all they avoid moot cases because of the lack of advocacy as a push-back.

Perhaps the most insidious idea is the concept of a network itself as a service rather than seeing the particular technologies as means to facilitate people doing their own networking. I’ve started to compare bit networking to social networking and providing networks is like arranging marriages. A key lesson of the Internet is that we can create solutions at the edge – even without a network as such!

We need a real marketplace that can be challenged to discover what is possible – this is the reason for replacing today’s Regulatorium-defined telecom with compositing connectivity from the edge. The example of the Internet demonstrates that this can work. And if we look at the post office that separates addressing from naming and thus greatly simplifies routing and we get rid of the perverse system of peering we can see how we can do very much better.

When I hear someone tell me what is possible – what the capacity is – I want to be able to say “let me try”. Today I have to limit myself to a choice of offerings from those who are afraid to try lest their business model collapses.


-----Original Message-----
From: craig@aland.bbn.com [mailto:craig@aland.bbn.com]
Sent: Saturday, April 19, 2008 16:02
To: dave@farber.net; bob37-2@bobf.frankston.com
Subject: how does one define "capacity"





[Dave F - for IP if useful]



Bob F asks:



    How does one define "capacity"? To use a trivial example -- what's the capa=

    city of a copper wire? We know it can usually carry a single phone call or =

    maybe 1GB Ethernet or ...



Which is an excellent question, but the way it is phrased here suggests

it is unanswerable.  However, it is answerable.



Fiber has three optical passbands in which light passes (relatively) cleanly.

The available capacity is about 25 terahertz per passband.   We've been

stuck at 1 bit per Hz for something like 20 years, so one fiber's capacity

is about 75 terabits.  We get far less today due to limitations of

transmission equipment -- so Bob's point applies that we can expect

to see continued improvement -- but somewhere around 75 terabits

(or 750 x 100 Gbps Ethernet connections) per fiber, we are forced

to use multiple fibers in parallel (a technique that fails somewhere

short of a yottabit [10^24] per second, at which point your

fiber bundles are the same diameter as subway tunnels, which seems

a good definition of implausible).



Copper is tougher, as there's crosstalk from other wires and other signals.

I'm not well briefed on copper, but when I talk with folks who do ADSL

signalling and the like, it is clear the basic rules are similar to fiber --

just with lower data rates.  Someone well versed in copper could probably

(for an ideal copper strand) do calculations similar to the ones for

fiber above for copper.



Summary: we can compute capacity.  We're not close.



Craig



PS: There's another version of this calculation that involves depreciation

schedules for installed multiplexing equipment, current ratio of terminated

and unterminated fibers in conduits, and ability to pull more fibers.

Don't go there :-).



*****************

Outreach Director, GENI Project Office

Chief Scientist, BBN Technologies

-------------------------------------------