How did a single computer failure take out the whole of the Melbourne rail network?

Going nowhere fast. Image: Marcus Wong/Wikipedia Commons.

If you don’t live in Australia, then your only experience of Metro Trains Melbourne is probably the ghoulishly jolly Dumb Ways To Die animated clip that went viral in 2012. But, as the clip loosely implies, the group operates one of the southern hemisphere’s largest metro rail networks.

Or, if you were trying to travel Down Under this Thursday, it turned out that they didn’t. A reported computer failure first led to massive delays on the whole system, and eventually to the entire central part of the city’s network being brought to a total halt for several hours.

With thousands of commuters stuck in trains, drivers and guards pleaded with passengers not to force the train doors and escape into the danger of the tunnels (possibly via jolly songs). A desperate Twitter user held a poll on whether he should answer a call of nature through the emergency door; naturally Australia backed him with an 84 per cent Yes vote.

Train failures happen everywhere: signal failures, track failures, accidents, and incidents of all kinds are common annoyances for commuters. But the whole network collapsing due to a computer error seems a bit much: could they not just turn it off and start again…?

Basically, no.

The biggest problem here is that Melbourne has a single underground loop line that links all its suburban tracks together. It’s a bit like London Underground’s Circle/District/Metropolitan/Hammersmith lines, with the crucial difference that almost all Melbourne’s suburban trains go at least partially round the loop – there’s no other route through the city. Even the city’s two terminus stations, Southern Cross and Finders Street, form part of the city loop, and many terminating services share signalling systems with loop trains.

A schematic of the Melbourne city loop. Source: Wikipedia.

Like many rail systems, Melbourne’s was randomly constructed by private companies out for a cheap buck, with no interest in interconnected metro or in buying up expensive city blocks to build through-lines. The city’s flat, swampy land also put everyone off tunnelling.

A lot of mergers and some nationalisation later, the system found itself feeding into two main terminals, both overcrowded. By 1969, transport planners decided that this wouldn’t do, and planned the City Loop.

Planners love joining up suburban lines with a tunnel through the city. You replace two busy, land-scarce, crowded terminals with through stations, serve major city destinations directly, and only have to turn trains around in the sticks. The RER in Paris, London’s Thameslink and Crossrail, and Berlin’s HBF all work on this principle.

But Melbourne’s planners were… braver.

Remember, every train into Melbourne already went into either Southern Cross (then called Spencer Street) or Flinders Street. Building the loop wasn’t like building Crossrail: it was like building a mash-up of Crossrail, Thameslink and the Overground to carry every single train that previously used to go into a London terminus.


The planners rose to the challenge: the number of tracks on the circle varies from two to six (originally planned as four to eight). But the net result was to create a complicated service with a lot of junctions and crossovers, which needs to carry lots of trains.

They had a clever solution to the complexity, too: they built one of the world’s first computerised railways. The City Loop opened in 1980 with fully automated signalling, using a dedicated system called Metrol, which allowed faster, more reliable switching, signaller and passenger information. When it opened, this was a state of the art system.

The trouble is, almost 40 years and some privatisation later, franchise operator Metro Trains Melbourne and government infrastructure owner Public Transport Victoria are still using it.

The brains of the system have been moved from the original ancient PDP-11 mainframes to a network of PC servers, and some modern train control systems have been added; but it’s still based around 40-year-old software, and the interface between the new computers and the tracks and signals still relies on 40-year-old interface boards. Metro Trains has to acquire replacements on the second-hand market as other companies retire their old servers.

Using old equipment isn’t rare in transport. New York’s still running on kit that was 40 years old when Metrol was installed, and the systems being removed in London are even older. Although the dreaded ‘signal failure’ is a common reason for scraping along at snail’s pace, they don’t halt the whole system.

But… well, have you tried watching an analogue versus a digital TV with a bad signal? Old-style train control systems are dumb, often fail, and are designed for the failures to be local and worked around. When newer-style train control systems fail, because the brains are centralised, so is the crash. More recent ones are designed to route around these problems, but Melbourne led the way and paid the price.

The complete system map. Click to expand. Image: Melbourne Metro.

Worse still, in other major cities, if a signal failure takes out one line, then you’ve still got a bunch of other lines to use. The City Loop means you can’t do that in Melbourne: if you lose signalling here, then every train line in the city is stuffed. Although the Metrol system only covers the City Loop and a few neighbouring inner suburban stations, you can’t run trains in the suburbs if they’re stuck on the wrong side of the city.

So what can Melbourne do? Replacing Metrol is often discussed, but would be painfully expensive, especially in terms of the disruption caused – and more recently, modern information systems have been built on top of it, which you’d have to fit in with the replacement.

The government are also starting work on a new tunnelled line through the CBD that's completely independent from the loop, which will help take some of the pressure off services – but even when it’s done, most city lines will still be subject to the same problems.

If I were Public Transport Victoria or Metro Trains, as the old joke says, I wouldn’t want to be starting from here.

Want more of this stuff? Follow CityMetric on Twitter or Facebook
 
 
 
 

What do new business rates pilots tell us about government’s appetite for devolution?

Sheffield Town Hall, 1897. Image: Hulton Archive/Getty.

There have been big question marks about any future devolution of business rates ever since the last general election stopped the legislation in its tracks.

Not only did it not make its way to the statute book before the pre-election cut off, it was nowhere to be seen in the Queen’s Speech, suggesting the Government had gone cold on the idea. (This scenario was complicated further recently by the introduction of a private members’ bill on business rates by Conservative MP Peter Bone, details of which remain scarce.)

However, regardless of the situation with legislation, the government’s announcement in recent days of a pilot phase of reforms suggests that business rates devolution will go ahead after all. DCLG has invited local authorities to take part in a pilot scheme which will allow volunteer authorities to retain 100 per cent of the business rates growth they generate locally. (It also notes that a further three pilots are currently in operation as they were set up under the last government.)

There are two interesting things in this announcement that give some insight on how the government would like to push the reform forward.

The first is that only authorities that come forward with their neighbours with a proposal to pool all business rates raised into one pot across a wider geography will be considered. This suggests that pooling is likely to be strongly encouraged under the new system, even more considering that the initial position was to give power to the Secretary of State to form pools unilaterally.

The second is that pooled authorities are given free rein to propose their own local arrangements. This includes determining, where applicable, a tier split (i.e. rates distribution between districts and counties), a plan for distributing additional growth across the pool, and how this will be managed between authorities.

It’s the second which is most interesting. Although current pools already have the ability to decide for some of their arrangements, it’s fair to say that the Theresa May-led government has been much less bullish on devolution than George Osborne in particular was, with policies having a much greater ‘top down’ feel to them (for example, the Industrial Strategy) rather than a move towards giving places the tools they need to support economic growth in their areas. So the decision to allow local authorities to come up with proposed arrangements feels like a change in approach from the centre.


Of course, the point of a pilot is to test different arrangements, and the outcomes of this experiment will be used to shape any future reform of the business rates system. Given the complexity of the system and the multitude of options for reform, this seems like a sensible approach to take. But it remains to be seen whether the complex reform of a national system can be led from the bottom up. In effect, making sure this local governance is driven by common growth objectives, rather than individual authorities’ interests, will be essential.

Nonetheless, the government’s reaffirmation of its commitment to business rates to devolution and its willingness to test new approaches is welcome. Given that the UK is one of the most centralised countries in the western world, moves to allow local authorities to keep at least some of the tax revenue that is generated in their area is a step forward in giving places more autonomy over how they spend their money. That interest in changing this appears to have been whetted once more is encouraging.

There are, however, a number of other issues with the current business rates system which need to be ironed out. Centre for Cities is currently working on a briefing of the business rates system, building on our previous work in this area, and we’ll be making suggestions as to how the system can be improved.

Hugo Bessis is a researcher for the Centre for Cities, on whose blog this article originally appeared.

Want more of this stuff? Follow CityMetric on Twitter or Facebook.