Thursday, August 16, 2007

Industrial Houston and packets/cars vs. circuits/transit

A couple quick pass-alongs tonight. The first is from a Joel Kotkin editorial in the Wall Street Journal on the myth of deindustrialization in America, which argues that manufacturing is alive and well in this country. A couple paragraphs on Houston:

In Houston, not only is employment in the energy industry up, there's also a growing manufacturing sector and an expanding port complex, which together have contributed to a more than 10% increase in jobs over the past three years. "Everything's now hitting on all cylinders," suggests Bill Gilmer, a Houston-based economist from the Federal Reserve Bank of Dallas. He adds that other parts of Texas — Dallas, Midland-Odessa, Corpus Christi — are also experiencing rapid growth, particularly in exploration and oil services. While engineers and geologists are at or near the top of the food chain, manufacturing compensation averages $80,000 — $20,000 more than in information and financial services, and more than three times that in retail.

Broad-based growth of this kind in the manufacturing and allied sectors is intimately tied to infrastructure. Houston has recently completed a major expansion of its port, with an investment of $2 billion. Its airport is undergoing a $3.1 billion upgrade, and an additional $65 billion in road and transit projects are being planned for completion by 2025. Such infrastructure investment should be regarded as critical to a regional economy, as Minneapolis is no doubt learning now.

The second pass-along is from Bob Poole's Surface Transportation Innovations newsletter at the Reason Foundation. The specific story is "Why Cars Beat Transit: Packets vs. Circuits", which I found to be a fascinatingly insightful analogy, albeit a bit nerdy. In its entirety:
Why Cars Beat Transit: Packets vs. Circuits

For months I've been meaning to write about an impressive paper from the Kennedy School of Government, "The Impacts of Commuter Rail in Greater Boston." It's based on a longer paper by Eric Beaton that was awarded the Howard T. Fischer Prize in Geographical Information Science, for the best use of GIS by a Harvard graduate student. In his research Beaton used GIS to analyze the relationship between rail transit station location and transit use. He found that "In the late 19th and early 20th centuries, commuter rail service played a major role in shaping the land uses in the communities it served. But that does not seem to be the case today. . . . Looking to the future, this means that providing new commuter rail facilities is not likely to produce significant changes in travel and land use patterns." (www.ksg.harvard.edu/rappaport/downloads/policybriefs/commuter_rail.pdf)

I held off writing about this paper because I despaired of its being able to persuade those not already convinced that rail transit has only very limited potential. So many elected officials, in particular, seem to have concluded that getting people out of their cars is (only) possible with snazzy new rail transit, so we must forge ahead, despite the enormous cost and very modest results in most cases. This seems to have become more of a belief than a reasoned conclusion. And against a belief system, what good is evidence?

But last week I read a commentary that provided one of those "Aha!" moments. Stephen Fleming, chief commercialization officer at Georgia Tech, wrote it for the Georgia Public Policy Foundation. Fleming uses the power of analogy, explaining the historic development of telecommunications on the basis of switched circuits. In "legacy" telecom systems, all the way through the failed ISDN of a decade ago, a voice or data message required its own, exclusive pair of copper wires, from origin to destination, with numerous switching steps along the way. But then a few pioneers invented packet switching, in which a voice or data message is broken up into individual data packets, with each one routed independently to its destination, making use (simultaneously with zillions of others' packets) of the entire network of possible connecting paths. Packet switching is what made the Internet possible: it's faster, more flexible, and ultimately much cheaper than circuit switching, though it requires lots of processing power.

As Fleming goes on to write, "The analogy is obvious: Mass transit systems are circuits. Automobiles are packets. Packet switching always beats circuit switching." He goes on to explain that mass transit means capital-intensive routes with minimal flexibility. But with automobiles, every "packet" goes into a large network to be routed direct to its destination. Especially when those destinations are door-to-door-to-door-to-door, as in so many people's trip-chaining commutes. (You can read Fleming's complete commentary at www.gppf.org.)

On second thought, perhaps this analogy is too abstract for non-techie elected officials. But it should be persuasive to the business leaders who end up on task forces devoted to solving urban transportation problems. Just remember: packets beat circuits.

Have a great weekend in (extremely) Tropical Texas...

Labels: , , , ,

13 Comments:

At 2:03 AM, August 17, 2007, Blogger Ian Rees said...

Seems to me walking/biking/rail is also packetized transit (and walking/bike has infinitely more routing redundancy than cars.) The problem is simply the scale of the network in relation to human ability... which is why I think cities need to be built on a scale that enables this (we've done it for thousands of years but forgot how in the past 50....)

Of course, it also seems people forget cars also need roads (routing..) to travel. Want to talk about that "series of tubes getting clogged" ? Freeway of your choice, 5pm. Enjoy.

But I suppose I'll see yet another useless analogy trotted out in every discussion from this point on...

 
At 3:14 AM, August 17, 2007, Anonymous Anonymous said...

Tory,

I got a good laugh out of this one!

1) Make sure your network doesn't get hit with a DoS attack. That holds true for your transportation network too.

2) Are we using TCP or UDP as the transport protocol? One is guaranteed to get to its final destination the other isn't. This analogy holds more or less holds true for your transportation network.

3) Speaking as a sysadmin who administers 2,500 machines in three clusters and is on call 24x7, it doesn't help when your user community keeps trying to do things that your network, storage, controllers, etc, was not designed to handle. That also holds true for your transportation network.

Enjoyed it.

Neal

 
At 8:44 AM, August 17, 2007, Anonymous Anonymous said...

I had fun with this. How about these examples?

- Do you need to implement a bridge to solve your network problem? How about a gateway, a router, or do you merely have to install a switch? This holds true for your transportation network.

- What kind of network do you employ, plain Jane Ethernet, or do you buy yourself something a bit fancier like Myrinet or Infiniband? Splurging a lot of money on a solution your users ultimately don't use will not get you much. This is especially the case because users have a habit of saying to you that they want a network that will solve certain problems, but they subsequently decide that they really want to do something else - like solve bigger problems.

This is really bad when your users are the people who are the ones that make money for the company, ergo you have no choice but either quit or to put up with their nonsense. They are the rock stars and you are the roadie. Their problems are your problems. This analogy especially holds true for your transportation network.

A minor aside above, DoS attacks are really implemented against servers, but they clog up networks in the process.

Have a great weekend!

 
At 12:15 PM, August 17, 2007, Anonymous Anonymous said...

Could it be that our port is expanding so quickly because we're importing so much from China (because our manufacturing is dead)?

The only problem with packets is, what do you do with all the boxes? Instead of 100 people in a box on the circuit, you have 1-2 people to a box in the packet model. And then when the people are delivered, all those packets have to be stored somewhere. Imagine what downtown Boston would look like if the only way to get there was by car. Half of it would have to be destroyed in order to build facilities to hold all the cars, and the other half would die from the resulting lack of synergy.

Or imagine the neighborhood around Fenway if suddenly the only way to get there was by car. It would all have to be leveled in order to build parking lots and parking garages. Instead of a great ballpark neighborhood, you'd have something like the area around Minute Maid.

 
At 1:17 PM, August 17, 2007, Blogger Michael said...

I agree with the analogy of rail vs cars to a certain extent.

However, in real life, your "packets" are 6,000 lb cars that are 10ft long * 7 ft wide * 5 ft high. And your "bandwidth" is limited - as you witnessed in California, and as everyone witnesses in Houston pretty much every day for 5-6 hours of rush hour. Meanwhile, a single rail line is equivalent to something like 6 lanes of highway in each direction, or 12 lanes total.

Also, 2 factors contributing heavily to manufacturing's success here:
- Non-unionized labor
- Steady supply of illegal labor

The question is: is that the model for the rest of the nation? It's good for business, and good for migrant workers, and some good jobs, but it doesn't replace what we've lost up in our northern industrial centers. I don't think those places can be replaced, due to competition from China.

 
At 12:34 PM, August 18, 2007, Blogger Ian Rees said...

Yes, the stretches the car apologists make for kind of argument in favor of an environment built for cars (as opposed to say, an environment built for humans) are astonishing.

 
At 12:35 PM, August 18, 2007, Blogger Ian Rees said...

(quick note: apparently there are two "ian"s commenting here. Amusingly we have similar positions. I should sign mine with something to distinguish.)

 
At 1:33 PM, August 20, 2007, Blogger ian said...

Oh geez, are there TWO commenting on THIS particular post already? Because I haven't added anything yet. . .

 
At 8:36 PM, August 26, 2007, Blogger Unknown said...

Interesting and though-provoking analogy. However, it's completely invalid in terms of transit for one critical reason. People don't commute randomly within the city.

Highways from the suburbs to downtown are actually like dedicated circuits themselves (as ian mentioned as well). But a train has MUCH more "bandwidth" in one lane than an equal-length set of cars. Most of the traffic in rush hour moves in one direction, along a spatially-limited number of lanes.

The only way that analogy would work in terms of transit would be if everyone lived and worked at two different, random points within the city.

Come on dude! You couldn't see that obvious gaping hole in that analogy? Really?

 
At 8:39 PM, August 26, 2007, Blogger Unknown said...

Also good points michael. I would also add that not only is bandwidth limited by the size of cars, but also but their top speed (~70 mpg at best in civil society).

 
At 9:46 PM, August 26, 2007, Blogger Tory Gattis said...

> The only way that analogy would work in terms of transit would be if everyone lived and worked at two different, random points within the city.

Actually, that's probably close to the truth. Houston has a dozen major job centers, and even all of them combined probably don't add up to more than a third of the metro jobs - meaning 2/3 are essentially at "random" points around the city.

 
At 12:05 PM, August 30, 2007, Anonymous Anonymous said...

Great, then that's at least 1/3 that could benefit from a rail system.

 
At 2:37 PM, August 30, 2007, Blogger Tory Gattis said...

Ah, but the people's homes are randomly scattered. You can theoretically lay out a commuter rail network to connect everything up, but then you realize that once transfers get factored in, people are looking at hour+ commutes each way - something they simply won't do. This is why I advocate express buses in a managed lane network instead: they can do the "packet switched" point-to-point at 60mph. The reason places like NYC and Chicago work is the gigantic concentrations of jobs they have downtown. This is not the case in Houston, where less than 7% of jobs are downtown.

 

Post a Comment

<< Home