System Overload | Hacker News Reader

System Overload

newyorker.com

59 pointsdonohoe15 comments

Two-track subways have inherent maintenance problems. New York is mostly four-track, so sections of track can be shut down for total replacement. They also have a huge number of track switches and interlockings, and can reroute onto alternate tracks easily. NYC can also run trains 24/7. Two-track systems can't do that. One of the legacies of the Tammany Hall regime is that the infrastructure from about 1900 was way overbuilt. But the real reason NYC is four-track is that the subways replaced elevated lines that were at capacity with two tracks.

Few other subways are four-track. The Broad Street Line in Philadelphia is four-track, but is underutilized. A few sections of the Chicago MTA are four track. The London Underground has the two-track problem. (Worse, that system is older and has undersized tunnels. They have to build special maintenance machines that will fit in their tunnels.)

In the US, there are probably two systems that could justify four-track - SF's BART and Washington's Metro.[1] But it took both a long time for each to build the traffic to justify it. BART has a lot of elevated sections, and four-track elevated lines roof over an entire street. (NYC used to have some; getting rid of them was one of the reasons the subway system was built.)

In practice, two-track systems do heavy maintenance during weekends and holidays. This takes careful planning, special equipment, and lots of competent, hard-working people. But it's done routinely by the big systems. SF Muni is shutting down their underground lines from 9PM to 2AM, and the Twin Peaks tunnel will be closed for 21 weekends starting in June. It's more expensive than a full shutdown, but has less impact on customers.

[1] http://greatergreaterwashington.org/post/3760/was-the-lack-o...

There was this great video last year, illustrating how outdated and error prone the signalling technology in the NY subway is: https://www.youtube.com/watch?v=Mjx3S3UjmnA

Hey, they're mostly General Railway Signal's NX system, the first intelligent user interface ever. It takes a huge number of big relays to express all the logic, but it's a good system. Everything in rail control since NX looks a lot like NX, even if it's on a screen.

Before NX, someone had to manually set up routes through an interlocking, setting all the switches and signals. The interlocking machine would prevent conflicts (the levers would not move when pulled) but the dispatcher had to figure out the routes. With NX (eNtry eXit), you push a button to select the track on which a train is approaching. Lights then light up for all the places it can exit the interlocking. The system takes into account other trains, previously set up routes, and even switches and track locked out of service for maintenance. You push one of the exit lights. The system then sets up all the switches and signals to get the train out of the interlocking at the desired exit. As the train progresses along the path, the resources (track sections and switches) are released behind it, and become available for other trains.

NX works well. Many later systems were worse. BART had big problems with their 1960s advanced electronic system, which had to be replaced.

(That video shows an older lever system. That's the only one on the NYC subway system.)

> Republicans have been less willing to support it, especially when it goes toward things like public transit.

I always thought that aspect of their modus operandi rather tricky and good. To demonstrate how government programs fail, they strangle them (cut their funds), then those programs start failing, and they later point out how government managed <whatever> is clearly a failure and it should be handed over to <somebigcorp>. Whether you agree or not with spending and the core idea, gotta give to the devil what is owed to the devil -- it is a pretty good PR strategy.

This is a common pattern that has a name and Wikipedia article:

https://en.wikipedia.org/wiki/Starve_the_beast

Ah nice find. Makes sense that it has a name.

Starve the beast is about forcing spending cuts by passing tax cuts first. People hate taxes. Plenty of people also use the opposite strategy of forcing tax increases by passing spending increases first. People love spending.

It has absolutely nothing to do with sabotaging the execution of public works project. You can't seriously think that Republicans are the cause of the graft, waste, and mismanagement associated with government infrastructure projects.

> You can't seriously think that Republicans are the cause of the graft, waste, and mismanagement associated with government infrastructure projects.

Most people think that it's the folks on the other side of the political isle from them that cause the problems.

Infrastructure problem costs too much? Is it the union construction workers that are inflating costs or the under-budgeted maintenance? There are a dozen convenient excuses that people of either party can blame on people not in their tribe.

Rah rah rah go our team, it's the other guys that are the problem!

> The tax cuts and deficit spending of former US President George W. Bush's administration were attempts to "starve the beast."

OK, so he was starving the beast in Iraq?

Just because our infrastructure is lousy doesn't mean "underinvestment" is the cause. However, you measure it, American infrastructure has become extremely expensive to build. Countries with nice infrastructure---like Japan and European countries---can also build it much more cheaply.

A number can be small because the numerator---the total amount of funding---is small, or because the denominator---the amount of infrastructure purchased per dollar---is large. In the US, it's mainly that the denominator is large. And because the denominator is large, we choose not to make the numerator as big as it could be. Also, the point about public transit is a real cheap shot against Republicans. Public transit accounts for a small percentage of total infrastructure investment. Even if it were doubled, this would have almost no impact on our infrastructure budget. And much of what transit is built now fails to pass a benefit-cost test. Look at BART's extension to Livermore. Look at the new Bay Bridge span, which now has a whole book about it http://www.citylab.com/politics/2015/10/from-250-million-to-...

Senescence and capacity are half the problem; the other half is that the teams (public as well as private) maintaining these systems don't have the logistics capacity to 'repair in place', to make substantial fixes without taking a train line completely offline.

(And I assume municipalities aren't offering to pay extra for that service).

This stuff costs money to maintain and upgrade.

I wonder if our whole culture's commitment to taking care of infrastructure has been eroded by the computer industry. We can just shut off the power on our old racks of servers, hand over their carcasses to the steel-recycling guys, and move their workload to AWS or someplace, for a net savings.

It doesn't work that way with bridges or railroads. It doesn't seem to work that way with air traffic control systems. Those things have to last at least a century, and have to be repaired in place, while still in use.

But when pols propose asking us to pay for that stuff, we refuse and vote them out of office. There's always some perfectly good excuse to say no: government waste, crooked contractors, expensive labor, you name it. Bottom line: we behave like the economic version of Moore's Law applies to highways. It doesn't.

Commodity computer hardware, to say nothing of AWS, hasn't been around nearly so long as our habit of skimping on infrastructure funding.

BART is famously underinvested and is struggling with the same issues. For example: https://www.bart.gov/news/articles/2016/news20160316-0

Good thing the administration started working shovel ready projects in 2009. Otherwise imagine how much less change would there be.

Reply on news.ycombinator.com