New York’s Subways Are Not Just Delayed, Some Trains Don’t Run at All | Hacker News Reader

New York’s Subways Are Not Just Delayed, Some Trains Don’t Run at All

nytimes.com

159 pointsmartinald158 comments

The MTA's problems begin and end at a weak senior management unwilling to standup to Cuomo's frivolous micromanagement and a transit union unwilling to modernize.

Politicians should be pushing for serious procurement and labor work rule reforms, otherwise the systemic managerial and operational deficiencies will ensure that the deterioration of service we see now will repeat it self in 5 to 10 years, regardless of how much money or technology is poured into the MTA.

London and Toronto have been able to modernize much of their transit systems in last 5 years and its not because they have more money. Once you fix the top everything else will fall into place.

Look what Andy Byford has been able to do in 4 short years at the TTC. His 5-Year Plan to modernize the TTC focused on transforming corporate culture and updating internal processes, in addition to new equipment. The results of these changes have been overwhelmingly positive with the TTC recently being named best public transit agency in North America and The TR Class of TTC subway cars in May having a MDBF of over 924,000 miles.

If your interested in getting involved with transit activism in NYC I highly suggest you follow @2AvSagas on twitter.

A lot of what affects the NYC subway is the principle of the "Sick System."

http://issendai.livejournal.com/572510.html

The system is too essential for a stop-the-world overhaul, but so far gone, that only a stop-the-world overhaul can part the clouds.

Without a hard flush of antiquated mission critical infrastructure, millions of turbulent eddies and vortices will ensure that the system is always afflicted by parasitic equipment and debris.

Any sufficiently complex system will always run in degraded mode, because the individual mtbf rates for distinct elements will compound. Even a stop-the-world overhaul can't fix that, it will reduce the failure rate but it won't get rid of it. The trick to running such a system is to assume it is broken and still figure out ways of providing continuous service.

Case-study, large raid arrays. Last time I worked with one, it was exactly as you say, it was always degraded, and disks were constantly being replaced and arrays were constantly rebuilding.

So you fix the server case by a share nothing design. Put all data on 3 unrelated servers and query the fastest one.

Not sure how you do this with subway tracks...

We'll just do the same thing. Call the three systems BMT, IND, and IRT.

Touche!

That raises the costs a bit though, if you actually had full service via 3 lines. Public transpo is already insanely expensive per rider over the 10-20 year term.

It's not the track that breaks down, it's what runs on it. I don't see a reason why the electricity or signaling infrastructure can't be redesigned to have redundancies built in. If the aim with modernization is to replace old signals by new signals and still have just a single signal, that is still inviting calamity.

Many of the world's oldest subways have solved this problem through the use of specialized technology transition teams whose sole job is to develop minimal downtime solutions to upgrading technology. I'm no expert on these technical teams...I merely know of their existence through my own personal obsessive curiosity about infrastructure. But at least in terms of having relatively modern performance and technology, Paris, Madrid, Hamburg, Tokyo, Berlin, Osaka, Barcelona, and even Moscow don't seem to have any problems. Some might run what appears to be really old equipment...but it runs on time without any significant downtime. And with the exception of Tokyo (which has severe overcrowding problems which exceed the natural limits of subways) and London (which has extreme geographic, geological, archeological, electrical, and real estate constraints), they have mostly never let these problems impact the daily service.

New York is probably the most easily upgraded of these legacy subway systems, due to the insistence of the original subway companies on triple tracking. Not only does the triple track allow for express trains, but it also allows for zero downtime maintenance during non-peak hours. The electrical systems aren't a barrier like those in some cities. Even Tokyo and London, which have worse constraints, have fared better. If technology transition teams exist, I haven't heard of them. Upgrade engineering appears to be outsourced. The costs that the MTA quotes for transition seem absurd. Once again, I'm no expert on the engineering involved, but the upgrade costs are approaching new construction costs in other countries.

We know we have a major cost problem in US transit infrastructure construction. But this is not just construction. We also have a problem with maintenance costs, operating costs, capital costs, etc. Everywhere you go in America, we have critical transit infrastructure that is hobbled by costs. Where is the cost management? Why can't we get system-wide cost audits? If annual audits are required of publicly listed companies, why isn't it required of public organizations funded by tax dollars?

I'm sick of the explanation that our problem is funding...its not. Our problem is mismanagement. When you manage your systems well, nobody has any problem sending money your way. The demand curve exists for government funding just as much as it does for markets. When you provide more for a given amount of money, you generally get more money. Give us better management.

> Give us better management.

It's gonna be hard considering how many independent and adversarial managers the NYC subway has.

Can New Yorkers make ballot propositions? If so it might make sense to make the MTA responsible directly to voters who live in the counties that are served by MTA. The current situation where it exists halfway between the governor and mayor seems like a governance structure where everyone can just blame the other person, similar to the governance structure dysfunction plaguing Washington, D.C.'s metro.

And part of the dysfunction of the DC metro is the lack of dedicated funding.

Accountability and power need to stay in the same hands. Being held accountable for a task you are powerless to affect is a powerful recipe for dysfunction. In context, funding is power - the power to make decisions about who gets how much funding and resources to improve the system. If you have funding, but no accountability, then you will simply redirect that funding to projects which are are actually accountable for.

Electing an MTA head, without giving that head dedicated funding, will be a political nightmare.

The subway of Paris usually has only two tracks. They have started making it automatic (and so they need to use a modern signalling system) and they seem to be successful at doing it without disrupting traffic too much: http://cometandnova.org/improving-service-through-automation... and https://en.wikipedia.org/wiki/Paris_M%C3%A9tro_Line_1#Automa...

I have a better metaphor for this: It's too big to fail.

Not a New Yorker, but what I heard was that the tunnels only have two tracks, and both are needed for daily operations. So you can't replace one when it wears out because of the service interruption. If they had made the tunnels wide enough for three tracks, they'd have a spare for when they do repairs on one of them. And give them a place to do a rolling-upgrade of old middle-20th century signaling equipment.

Actually, NY subway has 3 or even 4 tracks portions: https://en.wikipedia.org/wiki/List_of_New_York_City_Subway_l... I think that's the reason why the New York's subway can operate during all night, because maintenance operations can happen on the other set of tracks.

Two-track subways are the norm generally, at least in the US. In fact, New York's is unusual in that many trunk lines have more than two tracks, mostly separate express tracks so that express trains can pass locals stopped at a station.

(And it is possible to shut down one track of a two-track line for at least part of the way, if you're willing to run only one train through the affected stretch of track at a time. It may condemn you to fifteen or twenty minute waits between trains, but you can do it. The Washington, DC metro has been doing this a lot lately on weekends.)

Why not just do work on it at night like they do on the tube?

New York's subway system runs 24 hours, and many residents rely on late-night/early morning service.

Is that the official reason though? "The service cannot be stopped even for planned maintenance" doesn't sound like a reasonable constraint.

They have been doing shutdowns of particular lines for planned maintenance, but almost always with replacement service in place (either other nearby subway lines, or for more isolated lines, buses). And then there are major tunnel repairs, which can lead to shutdowns of over a year. (There's an upcoming shutdown on the L which a lot of people are dreading, as there is no good replacement service of anything like the same capacity.)

There is enough redundancy between portions of the 7th and 8th ave lines to permit partial shutdowns and yet five years in they still can't seem to finish their hurricane repairs.

Which is great but it sounds like they need to compromise. Does New York have night busses?

Yep. In fact they're ridden so infrequently at night, they have a request-a-stop program after late-night hours:

http://web.mta.info/nyct/bus/howto_bus.htm

Ha, that's how most busses in the world work, at least outside large cities.

> The system is too essential for a stop-the-world overhaul, but so far gone, that only a stop-the-world overhaul can part the clouds.

Hogwash.

The "problem" is that you may have to invest for a long time , continuously before you see the result. This is politically untenable, not engineering untenable.

I don't agree with your diagnosis for one simple reason: The Carnarsie line (AKA the L train). The L train is unique because it is the only line in the system that doesn't share track with any other line. It is well used but if it shut down for a day (or even a week) the effect on the system wouldn't be so bad. Yet attempts to upgrade the L line have still failed. How do you explain this?

That's not true. The L line is by far the most advanced line in the system because of the reasons you said. When I lived in Williamsburg, I remember coping with weekend shutdowns all the time for upgrades.

But the L now is (I believe) only one of 2 lines with modern signaling and CBTC (computer based train control), it was the first to have countdown clocks, and it runs with the least gap time (headway spacing) of any trains in the system.

That being said, it still is extremely overcrowded because of waterfront highrise development and the general shift to Brooklyn. This is why you're starting to see conversations about extending platforms (which for the underground stations is incredibly expensive).

And yes, I'm aware of the Sandy tunnel remediation work shutting it down for a year or so.

Still, the L is already upgraded (so is the 7 which also doesn't share much track). It's getting CBTC on the other lines with modern signaling that will be the hard jobs because of shared track.

IIRC the original plan was to fully automate the line. It failed because of the Union. CBTC is a compromise not a victory.

A related problem is the political pressure to keep the single trip price artificially low, under the excuse that it's on behalf of the working poor.

And yet the working poor buying single trip tickets, because they can't afford monthly passes, end up paying more per trip than someone who can afford the monthly pass. So the policy doesn't even meet the goal that the policy purports to shoot for.

And still another problem is that the system capacity is insufficient for demand. The design has reached it's scalability limit. We need transporters in NYC, or something, at this point. A continuous running conveyor-belt "people mover" like in airports might be more efficient.

>"Officials at the Metropolitan Transportation Authority acknowledged that the schedules were not being met, and they blamed overcrowding for the shortfall."

No, the overcrowding is the effect of the MTA not addressing massive increases in ridership over an almost 20 year period. The MTA likes to pretend publicly that the ridership boom came out of nowhere and blindsided everyone. However the increase in ridership has been occurring and been every since 1998.

"Between 1975and 1998, New York City subway ridership hovered between about 900 million and 1 billion riders annually. The net change between the annual ridership in 1975 and 1997 was about 100 million, an average increase of just over 4 million per year. However, 1997 was a turning point for the relatively stagnant ridership growth experienced by New York City’s subway system. In 1998,just after the introduction of the MetroCard, over 137 million more rides were taken than the year before. The boom continued for the next two years, during which time another 181 million rides had been added to the annual total. Over the next 15 years, the system continued to increase its ridership totals at an average of 25 million more rides per year. In 2015, subway rides totaled more than 1.7 billion."[1]

To make the current situation with chronically late trains worse there is no way to know when the next train is coming until you have paid for a fare and are standing down on the platform. Little good it does you then to find out that next train is not for 12 or 15 minutes.

The MTA could have put the electronic arrival boards up on the street level by the train entrance so that people could make an alternate choice such as to use a Citibike, taxi, bus or even just walk instead. The MTA decided to reserve that space to sell advertising however. Despite their press release you will not find the next arrival times on any of these. See:

http://www.mta.info/news/2012/03/29/digital-urban-panels-deb...

[1] https://wagner.nyu.edu/files/faculty/publications/State%20of...

SO MUCH THIS.

Fortunately, because the MTA has exposed the live schedule data, Google Maps lets you check this (sadly, it isn't 100% reliable: oftentimes a train will claim to have arrived somewhere, or even be departing from a station, while it's still getting there).

>"Fortunately, because the MTA has exposed the live schedule data, Google Maps lets you check this (sadly, it isn't 100% reliable: oftentimes a train will claim to have arrived somewhere, or even be departing from a station, while it's still getting there)."

Yes and the issue is that the MTA doesn't make the service disruption data available.

See: https://medium.com/transit-app/how-were-bringing-real-time-c...

Why wouldn't you make this data available? Why stop short? Most likely because it would expose just how many unplanned maintenance disruptions there actually are.

For much the same reasons as not operating all the trains on your schedule, I guess - apathy, incompetence, mismanagement, under-resourcing, etc. It seems odd to expect a transit agency that isn't good at operating trains to be great at providing data.

>>an average increase of just over 4 million per year. However, 1997 was a turning point for the relatively stagnant ridership growth experienced by New York City’s subway system. In 1998,just after the introduction of the MetroCard, over 137 million more rides were taken than the year before.

That is a huge change. Frankly, knowing nothing else I would almost be willing to guess that what accounts for the change is better tracking due to the new MetroCard system.

I think that's unlikely, prior to the card system you needed to purchase a token and drop it in the slot in the turn style so they could still account for ridership.

A more likely explanation is that Metro Cards simply made it a better deal since they now allowed you to transfer between the subway and bus on the same fare and they also offered riders unlimited 7 day and unlimited 30 day options.

See: http://www.mta.info/news-subway-bus-new-york-city-transit-me...

The biggest reason for these delays I've found over the years it the signaling system and headway spacing.

Blocks on a railway system are designed to provide enough space and warning so that trains do not crash into each other. However, when you use line side signals and CTC (Centralized Train Control) to dispatch them, you end up with a set of rules that create spaces for occupancy that can easily be longer than the trains themselves.

In the NYC subway, speaking as an observer as opposed to an insider, the signaling and dispatching requires certain distances between trains and headways (time between trains) to provide safety of 2-4 minutes. However, this results in the cascading delays as one train that dwells too long in a platform can have a stacking effect on trains behind.

For example, if your minimum block distance is 1000 feet, and a max length train is 500 feet, a stack of five trains waiting to enter a station would actually occupy 5,000 feet of track. This could stretch to previous stations along the line.

The best solution would be dynamic blocks and cab signals, where the system itself defines spaces occupied by trains and only allows a train into a space where it can occupy safely, but also can make the distances dynamic based on conditions. Cab signals would be required as you can't have dynamic blocks with line side signals.

However, to do this requires a massive investment in new equipment both line side and on the trains that would take years to implement, along with a massive amount of painful work that would cause even more issues while it is being set up.

Sadly, running more trains probably won't cut it as the system literally cannot handle more safely.

London Underground trains on the Victoria line are driven by computer and use a moving block system

this included the stations: you would have one train coming in the back as the previous train was still going out the front

apparently they special cased the stations to not do this, as this used to cause upset amongst the passengers!

Can't you also use manual train control during delays? I was once on a badly delayed train (we were second in line to the station behind a train with "police activity"). The operator pulls up to the train that's being held with maybe one inch between the bumpers and let's passengers off through only the front-most door. More creative thinking like this could help alleviate delays and increase passenger happiness.

I think overriding the signalling system should only be done in extreme circumstances.

If it becomes normal, it's too easy to make a mistake during the procedure, and cause an accident.

Its worse than you think. I can't find the source right now but IIRC the NYC subway signals still run on the vacuum tube system designed in the early 1900s. At the time it was a massive technical accomplishment. Every small arrangement of vacuum tubes is like a purpose built computer for signaling a specific section of track with speeds and spacing pre-calculated. It successfully put an end to high profile accidents caused by human error. Today you could do the same thing with a micro-controller for almost no money.

I used to take the 456 line on Lexington Ave. back and forth between Home & Work for the past 5 years. In January, the Q line on 86th st. opened, and although it takes me ~10 minutes longer each way, it is a MUCH more pleasant experience.

The 456 is a nightmare, and I can tell you first hand, that it is crowded to a point where it is inhumane (imagine people literally fighting for a spot to have the "privilege" to smell the arm-pits in their faces).

The Q line is well worth the headache, and the frustration of dealing with the 456.

I ride the 6 every week day -- luckily I don't go during prime rush hour, but even when I do, I don't understand why more people don't go to the end car. It's ALWAYS less full any time of day.

Somewhat similar story, except I am now biking pretty much full time. The trains aren't sufficiently reliable anymore.

The 456 is the reason I stopped commuting to work before 10. Taking the L from where I live today isn't much different. The trick I've found is that if the sign says there are two trains incoming that are within a minute of each other, always take the second train. People will self-sardine themselves on the first and the second with be mostly empty.

That was my biggest lesson of the L or the 4/5/6. If you wait, the next train will be much, MUCH less crowded.

Another trick is that if I'm at Union Square trying to get into Brooklyn on the L and the station is packed, it ends up being faster and more comfortable to take the L backwards to 8th ave, and then grab a seat on the Brooklyn bound L at the 8th ave terminus.

I use this technique when I need to catch a Queens-bound 7 train in Midtown. It's much better to take a walk over to Hudson Yards and board there.

This doesn't always work. In my experience, if you're near peak crowding areas during rush hour, the next train is usually just as packed as the first. And then the next. And the next. So you've just watched 3 trains go by that you couldn't even fit on if you tried and you've been waiting about 5 minutes. God forbid they get delayed, even for 2 minutes more. I usually just try to cut my losses and shove in as soon as I possibly can.

I lived on the L train for 7 years up until recently.

In my estimation, shoving onto the train just wasn't worth it. It makes the ride much more miserable for people in the car, and it delays the line even more. The second point is important; the train situation isn't going to get better by people being selfish and gaining a few extra minutes at the expense of everyone else's.

More often than not, the "second train" strategy works wonders. I found that when it doesn't, I'd have to wait for multiple trains anyway. Maybe once in 7 years did I skip a train with room and get punished for it.

> In my estimation, shoving onto the train just wasn't worth it.

In my ~17 years of living in NYC, I've never found it to be worth it trying to squeeze myself onto a train. The ride itself is awful and literally you're at most saving a few minutes at the expense of a terrible, terrible experience in the car. The next train almost always has fewer people on the car, and if that's not the case, the one after that most certainly does.

[Edit: I've also been fortunate enough during my time here that I could always afford to be a few minutes later to my destination, that is not always the case amongst those who take the subway]

>The train situation isn't going to get better by people being selfish and gaining a few extra minutes at the expense of everyone else's.

Doesn't that pretty much sum up the New York experience though? If everyone could not block the intersection, stand back from the baggage claim, and wait their turn then life would be better for everyone.

This phenomenon is almost comical with the NJTransit Northeast Corridor lines into NYC. They frequently cluster two inbound Northeast Corridor trains (all the way from Trenton) with the express coming 3-4 minutes after the local (having left Trenton 20 minutes later). Even on weekends, the local is frequently standing room only (usually totally packed) whereas the express will have plenty of free seats.

At Secaucus Junction (the last stop in NJ before NYC Penn), there will usually be a ton of people waiting for the train. Invariably, they cram onto the first train to come (which is already full) despite the big flashing notices on the timing screen that there's another train to NYC two minutes behind.

I've seen adults throw straight-up temper tantrums at being unable to get onto that train when there's an empty train following right behind.

I commuted my first year out of undergrad from Central NJ. At the time, it was very frustrating because NJ Transit is so unreliable. All the stresses of a given day could metastasize to a point where a transit nightmare could make you break.

Now, take adults who pay one of the nations highest property taxes, throw in confusion as to where their dollars are going (they thought infrastructure was definitely a bucket), and you have a perfectly good reason to throw a "temper tantrum."

Most NJ complaints, and attempts to speak with your local politician, or even a local agency, falls on deaf ears. Thus, what can these grown adults do when they are pretty entrenched into Jersey? It's frustrating. I moved out of NJ after spending my HS years, Undergrad years, and 1 year of work, and vowed to never give them another tax dollar because their government is filled with people who simply do not care about your problems.

Those taxpaying adults also manage to vote in people like Chris Christie, who decided the "infrastructure bucket" didn't include transit (and that the allocated transit funds would be better sent elsewhere).

Man, I wish that was the case in London. The central/victoria lines runs at 36tph (less than 2 mins wait between trains) but they are crush loaded full between 7am and 9am. Luckily we have crossrail coming to relieve the central line, which took less than 10 years to build in its entirety. Unlike the 2nd Av Subway which seems to be taking about 30 years (ignoring the failed start in the 70s).

Does NYC suffer from the "ghost train" phenomenon that plagues SF MUNI? The danger in waiting for a second train here is that there might not actually be one.

This is why it is a real human tragedy that the MTA is literally the worst transit system in the world when it comes to new construction. Millions of lives could made better if they could do capital projects in the time and costs that cities elsewhere are able to do them in.

In a perfect world there are four branches of government: Legislative, Judicial, Executive, Cranky Mass Transit Dictator

Well, we had Robert Moses and look how that turned out

He was more of an asphalt dictator who barely tolerated mass transit.

But I agree with your point.

He was the opposite of mass transit though. One can argue that if he had wielded his powers for good, that NYC would be much better off today.

I live in a much smaller city where my total commute each day is maybe 10 minutes to basically anywhere I need to go. How do you justify the long commutes? Do you pass the time somehow?

In addition to those who don't have reasonable options to move elsewhere for family and many other reasons, NYC (and Manhattan specifically) is filled with people who can't imagine living anywhere else. A lot like the Bay Area in that regard. Even those who live elsewhere (e.g. bedroom communities to the north) often have to take the subway once they reach the city by train.

All of this, by the way, is a topic that's rarely discussed when people are going "build up!" and otherwise densify cities. Without the rest of the infrastructure it's a recipe for complete gridlock.

Cities have multiple tools to handle building more infrastructure as density goes up. There's land value capture which is a tax developers (or new occupants) agree to pay on any excess value created by a transit improvement (i.e. chicken before the egg). Then there are development proffers, which requires developers to pay for certain public improvements (i.e. egg before the chicken).

In theory. In practice, as we see, building things like new subway lines take decades. More money certainly helps but build a few skyscrapers of residential housing over the next decade and you might get incremental subway lines to support that in fifty years or so.

It shouldn't take decades though. Perhaps a decade max including planning. Other cities manage to do that. It seems to be a particularly US problem that any transport project of capital size >10bn becomes a complete disaster.

Then invest in things like top-end cycling infrastructure. Copenhagen has pretty bad subway infrastructure (at present it's missing some important destinations, like the central station, the M3 and M4 will create a lot of important connections). But the bike infrastructure is top-notch and high capacity. In some regards they can create a very car-lite-city by focusing so much on bicycles.

The Q doesn't really replace the 6 though. I lived on the UES and worked around Grand Central so taking the Q would've probably doubled my commute. The 6 is a nightmare in the morning, but the Q actually helped with that. Anyone going to the west side would take the Q instead of the 6 to 51st to transfer to the E.

I only lived about 2 miles from work so as long as there wasn't a foot of snow on the ground, I would just walk. 40 minutes of fresh air beats 20 minutes on a stuffy train every time.

It was definitely a welcome addition though. It probably takes half the time to get to Penn Station from the UES now.

So the transit authority decided that it was more important to have trains evenly spaced out (to minimize the worst case wait times), but the tradeoff was less total capacity and thus more crowded trains.

I'm really not sure which is the better case. Trains always arrive in a short time, but can be full. Or you may have to wait longer for a train, and then a little longer still for the train right behind that train that isn't as full.

That’s a good summary.

This kind of problem is not easy to solve- it’s not the kind of thing that has a closed form solution. You typically use a discrete event simulator to model the outcome of different strategies. It seems clear to me, though, that the answer is going to be a mix of “maximize number of trains” and “space them out nicely” during rush hour. I bet the optimal strategy trades off those two variables based on some probability distribution of “expected number of available trains” over time, such that if the number of likely available trains is lower you space them out more, and so on.

its more than just that though. it's a legit safety issue. the switches are old and have been breaking down at spectacular rates. if a switch breaks when trains on the line are too close too each other there is a much higher chance for a collision due to one of the trains breaking too soon or too late. spacing them out more evenly reduces the chance of that scenario.

It might also be that "longer worst case wait times" risk turning into a different failure mode entirely. Like maybe if the rush hour train is 5 minutes late it's just uncomfortably packed, but if it's 10 minutes late it gets so packed that letting people off at each station now takes 3x longer than it should, the wait time spikes further, and every other train in the system gets stuck behind that one.

As someone with difficulty in crowded places I would have preferred less crowded trains over anything. Many times I had to wait until the evening rush was over before I could get on the 1 train home - the morning rush I had no choice.

I'm always impressed how terrible the subways are in America.

In Moscow the metro is running really smoothly since the day of it's creation except for a single day. But that was because of Hitler.

just in NYC because of its age. I don't think it's so bad in other big cities.

I don't care about the delays or capacity. I just wish NYC subway stations weren't so filthy. Yes, they've gotten better in the past 20 years, but every time I go to oversea, I'm reminded of how disgusting nyc subway is.

I honestly don't know how Moscow does it. I never saw a garbage can while at the same time never seeing any garbage laying around.

I think Japan does something similar.

http://jpninfo.com/54373

We're really really REALLY REALLY bad at infrastructure, for reasons that experts don't actually 100% understand yet. There's plenty of possible reasons that contribute, but most of them tend to boil down to some feature of our politics (and the process by which infrastructure construction is purchased). Looking at cost per mile of apples-to-apples transit systems is horribly depressing compared to pretty much any other country.

Why would it take 50 years to install signalling across the system ?

This doesn't sound right at all.

the only right way to answer that question involves a 10000 word history of the politics of the MTA, the finances of NY state and NY city, bankruptcies, budget shortfalls, political corruption and misdirection, very messy labor union politics, and the complexities of running a giant subway system 24/7 in a region that suffers significant damage every year from hurricanes and ice storms.

Unions.

Care to elaborate?

They're not typesafe, and that slows down refactoring.

The solution, of course, is to rewrite the subway in Rust.

As much as I think unions are an absurd way to deal with income inequality by doing everything terribly and expensively, this is a really unsatisfying answer when it comes to explaining why the US is so bad at infra construction. Freaking Germany and France are hardly famous union-busters, and somehow they manage to be infinitely more efficient than us at public works projects. There's clearly something more complicated going on than "Unions."

The problems with NYC subway signaling have been very extensively covered in previous reporting that you might find interesting:

https://www.nytimes.com/2017/05/01/nyregion/new-york-subway-... http://www.nydailynews.com/new-york/nyc-subway-relies-decade...

"Even something as basic as a cable is an antique.

Workers popped open a junction box to show a 70-year-old cloth-covered cable, due for the scrap heap next year, connected to newer rubber-covered wires."

Basically, the system is so old that an incremental retrofit is very complicated to achieve without full shutdown. Remember the overall NYC subway system is the only train system that actually operates 24-7 so maintenance scheduling is also much more involved.

> actually operates 24-7

Good one. They routinely shut down lines and stations for hours/days at a time. The L got shut down between Bedford & 8th Ave for the entirety of Memorial Day weekend. Even when lines are running perfectly, at night they only run once every 30-60 minutes. To riders the changes feel arbitrary and are impossible to predict. I've been stuck dozens of times in random places because I tried to use the wrong line at the wrong time. The only real way to know there's a problem ahead is to have the privilege of riding that line on a regular basis and reading the signs they post. If you look around almost every station will have posters denoting some change that's ongoing or coming up.

Actually, in much of the system there is a great deal of redundancy so one track can be shut down for maintenance using weekends and 10 PM to 5 AM. For example there are often parallel local and express tracks (6 local, 4,5, express or 1 local and 2, 3 express or C local and A express).

Also entire train lines frequently run in parallel a few blocks from each other.

See: http://web.mta.info/maps/submap.html

I'm not saying I haven't been caught inadvertently as well, but FWIW they actually do a pretty good job of posting all this information on their website.

And in fact Google Maps does a decent job of scraping it as well. You just need to remember that when Google Maps wants you to go to Jay St. Metrotech instead of High St it might not just be because it thinks it's 300m closer.

Then it should close at night. Pretty much every city with a subway system worth bragging about does.

It always seemed strange to me that Philly is the only city I've been to that fully replaces the subway at night with buses.

Part of what makes the subway system worth bragging about is that it runs at night.

I am proud of that as a New Yorker, but I would much rather replace it with buses at night if that would make it actually work during the day.

Why? Surface buses can run quickly at night when there is light traffic, and the air quality is far better at surface level than in the subway.

Even though trains are scheduled through the night, they shut lines down pretty frequently during overnight and weekend hours and replace them with bus service.

>Then it should close at night.

No, it shouldn't, and I take it by this statement you don't live in NYC. I do, and have lived in Tokyo and other cities with excellent mass transit. While Tokyo's system is much, much better, it's pretty fantastic that I know I can always get a train home no matter what time it is, no matter where in the city I am, much faster than any ground vehicle would take me.

Everywhere else just runs replacement buses at night and it's fine. Yeah ok it's a little slower, but it means you're not running a huge train for five people.

Many larger cities in Europe run the metro continually from Friday until Sunday night.

This is a decent compromise.

There are way more than five people per train at any time of the day or night. Besides all the people with night and early-morning jobs, there's also the fact that bars are open until 4 AM. They call New York "The City That Never Sleeps" for a reason.

What city in the US has a subway system worth bragging about? Philly's subway has 3% of the ridership of NYC's. It's not even in the same ballpark.

> NYC subway system is the only train system that actually operates 24-7

Chicago's subway is 24/7, too.

This is only true for the red and blue lines though. And since they only have one track for each direction they must interrupt the service to do maintenance.

It depends on where in the system the maintenance is needed. On the Blue Line they just have the trains share a single track for a while while work is being done on the second track. On the red line, they do the same thing outside of downtown. If the work is being done downtown, the Red Line trains can use the Pink/Green/Orange/Brown/Purple tracks and go elevated instead of underground.

Only the Blue and Red Lines are 24 hours IIRC

You are correct. But that doesn't negate the point.

First they have to make new rolling stock, as the new system only works on post-2000s trains.

Second, MTA estimate about 16 miles of track could be upgraded per year. They have 665 miles of revenue track (850 including non-revenue), so it would take 41.56 years just to convert the revenue track. The Regional Plan Association wants MTA to convert 21 miles of track per year, leading to 31.66 years to cover all the revenue track.

Add the time to create new rolling stock, any additional time for any additional track in the next 30+ years, and extra time in case they go over the original estimation. Easily between 35 and 50 years just to convert the revenue-generating track.

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

> "as the new system only works on post-2000s trains."

Actually, it isn't even all post-2000s trains. The CBTC [2] train controls aren't even on all of the post-2000s trains. In fact besides the 'L' train which has already been converted, only the 7 line has all of its cars as CBTC [1].

This is really bad news, because ideally, the very, very crowded 4,5,6 lines mentioned in the NYT article should be the lines to have the signals computerized, but the 4, 5 trains are relatively "new" having been put in service in 1999-2003 or so. Some of the trains in service are > 50 years old.

[1] https://en.wikipedia.org/wiki/New_York_City_Subway_rolling_s...

[2] https://en.wikipedia.org/wiki/Communications-based_train_con...

475 stations, 840 miles of track, the signals have to be 'hot patched' as the system still needs to keep running 24/7/365.

50 years might be an over estimate, but projects like these tend to slip their deadlines.

[Edit] peterwwillis's answer is much more thorough than mine https://news.ycombinator.com/item?id=14948951

Just implement speed restrictions (say 5 M.P.H.) in block sections that are being upgraded. Sure it's annoying to have your train slow down, but if you're going that slow you don't really need signals, to begin with.

We're underground. Is there space for the maintenance staff to stand whilst a train passes? Probably not everywhere.

If trains normally average, say, 30mph, and 12 times per hour in one direction, a 0.5mile restriction adds 5 minutes to the journey time. Now we can only run 11 trains per hour -- which messes up the whole schedule.

The periods where there isn't a train passing will be quite small. Is there time to get any necessary equipment onto the track, start the work, then clear the equipment away before the next train is approaching?

What happens if the staff make an error, and break the existing system? It might not be possible for trains to move, in which case we now need to evacuate hundreds of passengers along the tunnel.

Work within metro tunnels while trains are running is, therefore, very unusual.

Ozrtm

1) Much of the system is very old, some of it dating back to the 30s. Replacing those parts is very complicated, it's not a 1:1 process. The improvement comes because new signals work under different assumptions. The current signals are basically automated versions of people standing with flags. Partially interfacing the old with the new is very complicated.

2) Downtime basically shuts down the city. Even nights and weekends interruption in service causes major disruptions. The surface streets are at maximum capacity, there is no way to get that many people in and out of downtown and midtown without the subway.

> Why would it take 50 years to install signalling across the system ?

Because currently, the residents would riot over the taxes/fees cost of redundancies required to enable faster maintenance and upgrade projects of a 24x7x365 operating system, where the pain of maintenance is constantly deferred until it turns into a looming, critical safety-liability issue. Back when the population density was lower and the pace of life slower, and less economic dependencies upon this infrastructure, the disruptions were affordable and more palatable, but now there are likely powerful stakeholders pressuring the Port Authority from taking as many outages as they'd like these days, comparatively. When I used to live in NYC, the Port Authority took a lot of abuse from a lot of different quarters for every line shutdown for planned maintenance. Increases in fares to fund future expansion/maintenance was met with a lot of resistance. They were like sysadmins for NYC.

This is the real reason for all of these problems. It's political. The fares don't come close to covering the operating and capital cost of the system. It's not user-pays, it requires constant government subsidies (as an implementation detail, the required government subsidies form a convenient way for the state and city to hold each other hostage, but that's another story).

Any talk of use pays is met with outrage about how low income people out in the boroughs couldn't afford it, so the whole system continues to operate as a horribly inefficient welfare system for poor people to get to underpaid jobs in the city, instead of a transit system. It would help a lot to double (triple?) the fares overnight, and give the poor people out in the boroughs a free metrocard - at least that way you're capturing the value provided to the rich people and tourists all the way through manhattan and brooklyn.

Why would it take 50 years to install signalling across the system ?

Caltrain is on year 5 of their "3 year" project to modernize their signaling system... and that's a relatively simple 75 mile system with 30 stations and it's almost all above-ground.

It's not right, and the explanations about the age of the systems are mostly excuses. There is a good article somewhere, maybe the Atlantic, where the reporter was able to get some MTA managers to admit they were essentially holding fixes to the visible problem (no arrival time signaling) hostage as a means to gather public support and funding for less visible overhaul of the infrastructure, by bundling the two together.

Hard to prove this is the case, but it smells right to me. Modern signaling system that did nothing but give arrival times to the public would not have to make use of any infrastructure at all -- it could run completely independently -- and could be done for a tiny fraction of the billions of dollars that the MTA is spending on the slow work they say is currently necessary to get it.

Relevant article on a tax to fund the changes https://www.nytimes.com/2017/08/06/nyregion/bill-de-blasio-w...

Anyone know whether anyone's carefully considered simply adding new parallel systems, e.g. for signalling?

I can completely sympathize with the difficulty of upgrading 24/7 critical infrastructure, especially given that mistakes can kill or injure people. I'd try to just deploy separate parallel systems instead.

At a comically reduced scale, businesses migrating between ERPs (or similar line-of-business apps) often run both the old and new systems 'side-by-side' for some period of time. What I've never seen or heard them doing is trying to replace components or sub-systems of the old system with those of the new system. That kind of integration work would be awful and would probably drastically increase the odds that the migration would crash and burn.

When NYCT overlay's CBTC signalling on a line the block wayside system is upgraded at the same time and the two systems work in tandem. This allows non-CBTC equipped trains to operate over CBTC controlled territory and the wayside system functions as a backup should CBTC fail for any reason.

As you point the integration work for this is complex and is usually the primary driver of delays and cost increases.

This is what's actually happening to a large extent.

While CBTC signaling is being installed, the original fixed-block signaling remains in place. It's nearly impossible to do this otherwise - you'd have to shut the entire line down for the entire time it takes to install new signals - which could be years.

Once the new signal installation is complete the two systems may operate in parallel (with the fixed-block system acting as a possible backup if CBTC fails, and also so older trains can still run).

The issue is that with 24/7 operation, there is no consistent regular time where work like this can be done. It can only be done in short spurts (late-night shutdowns) or the work crews in-lined with train service, which has obvious safety concerns as well as drastically reducing the capacity of the line.

I actually think this is the wrong approach. You don't get to reap any benefits until you start eliminating the old system, and with the parallel approach you defer that cleanup period until the transition is complete. It'd be far better if adding CBTC yielded the cost savings of eliminating the old system now as it rolled out. That requires more complex inter-op systems and more expensive upgrades to trains, but both of those things are not the constraint. The core constraint is the time you can shut down the subways for in order to do track & signal work in place. When you have a critical constraint, you should do lots of extra work in other places to maximize the productivity of the constrained operation.

Is this just a problem with NYC's system, or is this an "elephant in the room" problem with mass transit?

If an individual's car breaks down, they can find alternate means to get where they need to go. If everyone were to depend on public transit in the future, how do you handle upgrades? I know the metro in DC will run extra bus routes for some down trains, but I think that's just for certain smaller repairs.

It's something I haven't really thought much about before. The simple answer is to have redundant equipment (tunnels, tracks, cars), but few would stomach the cost. Most likely it would just get used instead, in the same way that you can never alleviate traffic jams with more roads, you just get more people using them.

Almost every Subway supports transparent upgrades during operation. In the absolute worst case scenarios you shut down single stations at night. By design almost every Subway is redundant - you at least have to have two lines, one for station bypass and one for returning trains. When you work on any Subway, you shut down one of the lines and manage traffic through the other.

The NY Subway is just the most incompetent, most mismanaged, most technically indebted subway on Earth.

There is no "problem with mass transit" as much as it is a problem with centralized administration of complex things, period. Anything of large scale that impacts many lives but has limited controllers to manage it is both rife for ineptitude / corruption / incompetence and easily ignored on the consumer side because of the complexity.

The same issue turns up all over infrastructure - outdated bridges, power lines, internet infrastructure, water pipes. Central authorities of incompetent taxpayer dollar rent seekers run them, the people are too ignorant / ill informed about deteriorating conditions to care, legacy technical debt builds up over centuries to be impossibly large.

And you won't escape them. The inverse of your scenario - rather than the car breaking down, the road collapses from age - is still wholly applicable. Car travel and vehicular routes are just as mired in corruption and bureaucracy making them a nightmare as trains. Trains breaking down is as easily solved as cars - there are maintenance stations, you take one train off, you put a new one on. The real problem is always when the tracks go bad - the centralization part. You could have a billion trains but not enough track / switching capacity to use them.

In much of Europe, metro systems will shut down for maintenance for appropriate periods. This is massively cheaper than working around a live railway (especially underground), where safety would necessitate many more staff required and more interruptions to work -- if it's even permitted in the particular location.

The period depends on the city. Here in Copenhagen, where the metro runs all night, they tend to shut part of a line from 00-05h on a Monday-Thursday night, and put on buses instead. Cities with bigger metro systems might not need the buses: this was the case the last time I visited Berlin, and a central section was closed overnight.

London sometimes closes parts of lines at weekends, and every 2-3 years or so there is major work on a line (closed entirely) on ~25-30th December.

(To be clear then: the "absolute worse case" is closing an entire line for around a week.)

Balance the inconvenience of service interruptions with the cost of reducing the interruption.

Engineers and other staff can work at weekends / nights / popular vacation times / public holidays, but at an increasing cost. There's also a limited total number of engineers, so it will help to have a steady amount of work (i.e. permanently employed staff).

No, NYC's subway is particularly messed up for unique political reasons.

They don't have these problems in London, Tokyo, Hong Kong, Singapore.. it's not a problem with subways, it is a problem with the MTA.

> It's something I haven't really thought much about before. The simple answer is to have redundant equipment (tunnels, tracks, cars), but few would stomach the cost.

Given that your premise is that everyone is on transit, presumably some of the incredibly gargantuan subsidies that are handed to private vehicle drivers would be used. I don't know if it's the just-world fallacy or what, but people always seem to have trouble grasping truly how inefficient car travel is. In a situation where gov't doesn't heavily subsidize them (and the drivers aren't paying the high costs associated with cars), I can't imagine that the cost of redundancy equipment required for no-downtime upgrades would be un-stomachable.

On the other hand, the other thing I know is that the electorate is far too stupid and far too selfish to grasp simple concepts like this: my own city has to fight tooth and nail against car drivers as a political bloc in order to claw back a fraction of the outsized privileged position they have in the transit infrastructure. So you may be right that it won't be politically feasible despite being less expensive than the alternative.

NYC MTA needs billions of dollars to fix the system. NY state must raise taxes on the rich and deploy it too the MTA. The federal must chip in too. The MTA will never make money and this is how it is.

Where does the misguided notion infrastructure should make a profit even come from? Everyone hates on Amtrak for not being a profitable company. When in history has infrastructure ever existed for the sake of the network operators profit? It is always for the sake of profiting those who use it - the infrastructure enables profits, but in the same way sticking profits into medicine cripples the efficiency of the system, trying to stick rent seeking into infrastructure defeats the point of using it.

Hong Kong has one of the best subway systems in the world and it generates a profit every year [1]. The USPS also generates profits as well [2]. I would argue that the profit motive incentivises efficiency in both of those examples.

The best innovation often happens when working under constraints; often a drive for profit results in better service for everyone using the infrastructure. Of course this isn't the case 100% of the time, like your example of medicine, which is why I think that to get the best of both worlds you need a mix of government regulation with privatized capitalism. But I think it would be incorrect to dismiss the profit incentive entirely as being bad, and probably in the case of US infrastructure we would benefit from more privatization and not the other way around.

[1] http://money.cnn.com/2015/03/30/news/hong-kong-mtr-subway-pr... [2] http://www.newsweek.com/post-office-aint-broken-its-profit-s...

That was USPS's first profit in 5 years just so we are clear.

http://www.govexec.com/management/2016/02/usps-records-first...

There's a couple interesting issues at play. One is that a profit incentive can be at work without the goal being to break even. For example, perhaps the goal is to subsidize by only $XXX million/year and above/beyond this goal is considered "profitable" with appropriate incentives.

Another is the issue of correlation/causation. Well-run systems may tend to be efficient in their use of monetary resources (almost by definition), but it's not clear that setting a for-profit goal will be a cure-all intervention that causes a system to be well-run.

It's also a big leap from setting profit goals to full-on privatization, which I'm skeptical of for several reasons. Private companies are often only pushed to be "good enough", better than the competition. A public transit system that is just barely better than a cab (say for daily work commute) is no good at all. Private companies tend to have short-term incentives, for instance, the CEO may get a bonus based on the current year's profit, which will be destroyed by investment in infrastructure. Companies care about maximizing revenue rather than welfare. Etc.

> When in history has infrastructure ever existed for the sake of the network operators profit?

Most railways in Europe and North America were built for the profit of their investors, and made a profit for a long time.

Before that, so were some highways, the canals, and many bridges.

In Britain, the railways were bankrupted by WW2, and afterwards consolidated into a national company (British Rail).

I do agree with your sentiments entirely but the MTA should be doing much better.

Transport for London (IMO the complete opposite of MTA: on the whole incredibly well run and generally well funded) has went from making an enormous operational loss to near break even in the last 15 years. If the trends hold it is projected to make a >£1bn operational profit per year by 2021.

NB I'm talking about operational expenditure not capital, which is funded from grants from govt, which currently stand at around £3bn per annum. It's possible to see how it could be entirely profit generating in a couple of decades though Inc capex. Assuming everyone doesn't leave due to brexit.

The MTA spends about twice as much per ride as the London Tube. It doesn't need more money.

The line about overcrowding being one of the major drivers behind delays doesn't hold much water. Ridership has essentially been flat or even declining in some areas[1] (peaked in early 2010s). Decreasing mean distance between failures in cars [2] (as well as mismanagement of crews & trains as mentioned in article) are among the main culprits.

[1] http://web.mta.info/nyct/facts/ridership/ridership_sub_annua...

[2] http://web.mta.info/mta/news/books/archive/160725_1030_NYCT.... <-- See page 29

Are you sure? nytimes recently said its doubled. (thougcomparison to 40s is interesting)

> Subway ridership has risen dramatically since the 1990s, when about four million people used the system each day. Now it is nearly six million riders each day — the highest level since the 1940

https://www.nytimes.com/interactive/2017/06/28/nyregion/subw...

They said it's risen since the 90s, which is true. That chart in the middle is a bit misleading (not granular enough). Scroll to the bottom of my first link to see the MTA's own figures on annual ridership for the past few years, you'll see it's down .3% since 2011 (although has gone both up and down in the intervening years). And just for the sake of completeness, I'll counter your NYT article with one of my own ;) [1]

[1] https://www.nytimes.com/2017/02/23/nyregion/new-york-city-su...

What caused the spike in the 1940s? Wartime fuel rationing?

The spike was in 1946. Whenever this statistic is trotted out I always wonder how they made the trains run back then but MTA can't figure out how to do it today. Their handwaving about narrow doors slowing down passengers doesn't hold water.

That's interesting. By 1946 most soldiers would have been back in the states but automobile production would still have been ramping up. New car designs didn't really hit the market until after 1948. Maybe that contributed?

Instead of cancelling trains, why not take the unused cars and make the running trains longer? I understand that the stations may not be able to handle the length. You could stagger entries so that the first half of the train loads, pulls forward a bit then the second half loads.

If it is an issue with the strength of the train engines, could you use two train engines together? Especially if their intended routes are just getting cancelled.

The trains on the 4/5/6 lines are already the full length of the platforms. (As on every other NYC subway line except for the G and a few of the shuttles.)

I was thinking that you could have trains that are longer than the length of the platform and have multiple steps to fill it.

It could be a safety thing. In case of an emergency you want to be able to just roll into a station and evacuate the train as quickly as possible.

Anyone know of a documentary on the New York subway? I would love to see all this old equipment and what they are really battling in order to keep it running.

There is this[1] video from the MTA that goes into the signaling system, which is extremely old and in the process of getting updated in certain sections.

[1]https://www.youtube.com/watch?v=Mjx3S3UjmnA

There are some pics in a recent nytimes article. Looks incredibly old, especially if you compare all the computing power in the phones of its riders.

https://www.nytimes.com/2017/05/01/nyregion/new-york-subway-...

The people in NY/NJ area have no choice but to endure the mess of NJ transit and MTA. Politicians lack vision and will to build a modern transportation system that accounts for growing number of riders through these systems. Somebody must have told try these folks to keep the "lights on" and forgot to mention that it also includes getting riders on time.

It would seem one could analyze the performance in this day of big data and find a more efficient plan for scheduling the trains and adjusting in real time. Maybe the system is too unpredictable to make this possible. Even so it would be fun to try to model the entire system and try.

The problem isn't in modeling the perfect train distribution, it is in that the system itself can't handle anything else. Many of the parts in the subway infrastructure are from the 40s and they just can't handle more trains.

> from the 40s

30s! And maybe even older!

Does anyone know how much those ridiculous touch screens cost?

You mean the ones that allow them to show multiple ads, changed digitally, without paying someone to manually change them out?

If those are net-costing anything at all, I'd guess it's a rounding error on any subway upgrades.

The worst part about this is the New Yorkers who are fleeing New York City and landing in Hollywood and Silverlake and Echo Park and Eagle Rock. No xenophobia here it's just they're bringing their adjusted economic expectations to LA making the Airbnb inflicted housing crunch seem OK.

I find it quite shocking how little of the NY subway is automated.

What would incentivize it to do so? Upgrades come effectively from the state, and need to be passed in budgets. So you need to persuade opinionated Albanian politicians to upgrade the NY Metro.

I know this would probably be unpopular but why not just temporarily raise prices until fewer people can afford the train. And use the extra money to upgrade the system.

In current ecosystem, with so many politicians promising golden eggs, price rise will almost never end up in the decrease.

Also, you are taxing the poorest. I think they struggle enough. You could start selling premium services though.

Would it be anymore unpopular than funding it in a way that makes sense? The subway is the most efficient way to move people around: if you shunt people away from it you may very well end up making your net financial position worse. At best, this amounts to accounting trickery.

Reply on news.ycombinator.com