Spy Plane Fries Air Traffic Control Computers, Shuts Down LAXA relic from the Cold War appears to have triggered a software glitch at a major air traffic control center in California Wednesday that led to delays and cancellations of hundreds of flights across the country.
On Wednesday, a U-2 spy plane passed through the airspace monitored by the L.A. Air Route Traffic Control Center in Palmdale, Calif. The L.A. Center handles landings and departures at the region's major airports, including LAX, San Diego and Las Vegas.
The computers at the L.A. Center are programmed to keep commercial airliners and other aircraft from colliding with each other. The U-2 was flying at 60,000 feet, but the computers were attempting to keep it from colliding with planes that were actually miles beneath it.
Though the exact technical causes are not known, the spy plane's altitude and route apparently overloaded a computer system called ERAM, which generates display data for air-traffic controllers. Back-up computer systems also failed.
As a result, the FAA had to stop accepting flights into airspace managed by the L.A. Center, issuing a nationwide ground stop that lasted for about an hour and affected thousands of passengers.
Please don't say they're storing height in 16 bit integers.
16-bit unsigned integers, I"d guess. Some smart person figured that you wouldn't see a lot of negative altitudes.
I hope they didn't sell that software to Atyrau Airport.
Flight levels are recorded with a 100ft LSB but tracked to only 500ft increments. You would have to be packing down to a 7-bit field to get an overflow in the right range: one bit for the hundreds leaves 6 bits for the thousands so overflow at FL640.
Since only 500ft increments are used, Atyrau Airport would round to FL000
That's the letter U and the numeral 2.
These guys are from England and who gives a shit?
ERAM is a black hole that the FAA keeps pushing money down to try and replace a 40-year-old outdated system. They went through the same thing in the 90s with a program called AAS, where they burned through almost three billion dollars before saying "fuck it" and going back to a system running on vacuum tubes.
ERAM is something like four years behind schedule, and they shovel twenty or thirty million dollars a month into it, only to have shit like this happen.
More fun reading:
http://www.baselinemag.com/networking/The-Ugly-History-of-Tool-Development-at-the-FAA/
http://www.ainonline.com/aviation-news/aviation-international-news/2013-10-02/eram-development-reminiscent-failed-aas-program
Why do people insist on fixing things that aren't broken? New Yorker magazine on the use of paper in air traffic control.
Bug... what bug... like y2k bug....
Probably worth reading this William Langewiesche piece as a side-order. http://www.theatlantic.com/magazine/print/1997/10/slam-and-jam/305134/
(PS, oh, it used to be so nice that your OpenID implementation worked. now it doesn't seem to provide a callback to the referring post, so just leaves you at the OpenID server.)
Beats me. Nothing has changed.
I admit that feet > 65535 was the first thing I thought of too. But yeah, since altitudes are stored /100 that's not plausible. The real explanation turned out to be something complicated involving 'VFR above clouds'. I guess the interesting question is what was different this time, why had it not happened on any previous U-2 flight.
Oh wait, I know: it was the first time there were clouds in Southern California.
Was it the first time a U2 actually filed a flight plan, or had their transponder on?
No, Nasa flies their U2 out of Edwards regularly, often heading out past LA to do flights over the ocean.
comp.risks
Interpretation of press comments:
Lesson learned: Don't put flight plans without altitude information in to ERAM. That will solve _everything_.