FAA using mainframe computers from Philips

The federal government does computing a bit differently than other folks. The FAA has a computer system for keeping track of a few thousands of flight plans every day and distributing those flight plans out to air traffic control centers. The current system went into service in 1988, according to the story linked below.

What did the FAA choose back in the 1980s? Keep in mind that 1988 was 10 years into the ascendancy of the DEC VAX line of minicomputers. It was 9 years after the release of the first version of the Oracle relational database management systems. It was 6 years after Sun Microsystems began shipping its popular Unix workstations.

According to this story, the system launched in 1988 on a pair of Philips DS714 mainframe computers (photos). Philips quit the computer business shortly after these machines were installed. This article from 2005 talks about the FAA’s plan to replace the Philips mainframes and to “go online early next year [2006]”, but apparently they haven’t completed the porting project. The volume of traffic is larger than you’d expect, with 1.5 million messages per day (about 20 per second; roughly what a pizza-box Web server might be expected to handle), apparently because the network handles weather and NOTAM data as well as flight plan data.

5 thoughts on “FAA using mainframe computers from Philips

  1. Most of the daily computing transactions people incur are still shuffled through mainframe systems, the core of which are older than the FAA system you quote. Whether it be charge card submissions, insurance claim adjudications, flight reservations, power meter usage, etc.…, odds are good that it’s core processing occurs on a mainframe, using an old school hierarchical database platform like IMS (though in many cases DB2 is running but no fancy SQL as the data structures are primarily in “flat file” format”).

    Not that these companies haven’t embarked upon reengineering and rewrite efforts, but most of them are years late and can’t even handle the small fraction of data they attempted to siphon off the “dinosaur” systems that still run.

  2. The Philips was never a mainframe-class machine. Instead, it is an old SITA-created packet switching machine that has remained technologically in the 1980’s.

  3. 20 messages per second? Depending on what the messages are and what it has to do with them, I think that’s what a pizza-box web server from 1996 might be expected to handle. Say, a Sparc 5.

    I think you can get 72MHz ARM7TDMI microcontrollers, coincidentally also from Philips (although spun out into a company called NXP in 2006) that should have performance similar to a Sparc 5, while running on a single AA battery. Like the LPC2888, or the LPC2468, with a clock crystal, an external SDRAM, and an Ethernet transceiver. (The LPC2468 has the Ethernet MAC built in.) They cost about US$10-$15 from Digi-Key.

    I guess we have reached the level where hardware is free. It’s software that costs money.

  4. Former colleagues of mine (they’re all aged around 60) were involved in designing, testing and production of these beasts. Apparently there were several very large installations of these “Dutch” machines in the US, from what I hear the largest one was located at Western Union in New York. There were computer floors containing hundreds of these cabinets in a building on Broadway with the Philips logo’s facing the windows.

    SITA was indeed the largest customer and had sites with these machines all over the world.

    Anyway, the business unit that produced the DS714 (and related products) went out of business a long time ago, some remnants remain here and there though.

    There are stories of a machine at Royal Mail in the UK which had no downtime for 20 years….

  5. Is there an archive of all the flight plans filed? It could be used in litigation problems and reconstructing information if logbooks are lost or destroyed.

Comments are closed.