6 ms·

Why exactly does this author think an airplane needs a lot of computing power? Airplane software does not deal with especially large amounts of data or intense

Why exactly does this author think an airplane needs a lot of computing power? Airplane software does not deal with especially large amounts of data or intense computations. Faster hardware does not provide an inherent safety benefit. There is a large degree of coupling between avionics hardware and software amd updating that software is prohibitively expensive. There's just no point in doing it unless you get a tangible benefit. Furthermore, avionics systems take a long time to develop are not often rebuilt from scratch, so the hardware naturally lags far behind the computing power we're used to. The recycling of avionics systems is definitely fueled by a desire to reduce costs. However, it's also important to note that older, mature systems with a history of service inherently offer a degree of safety compared to something which is untested. The issue this article complains about can easily be seen as a safety feature. To be fair, the author does spend most of the article talking about missing features and design issues rather than harping on "processing power". It is a poor title.

dmix 8y ago
Wait until they finds out what software powers the nuclear weapons of 90%+ of the world, top-tier fighter jets, "critical" infrastructure like power (including nuclear), and countless financial/banking systems. To be fair the article doesn't critique using 1990s (or earlier) hardware/software. It's merely a clickbait headline. The only reference is dismissed by the very next sentence as insignificant [Boeing could be replaced with any ~~expert~~ somewhat-knowledgable person on the subject]: > The flight-control computers have roughly the processing power of 1990s home computers. A Boeing spokesman said the aircraft was designed with an appropriate level of technology to ensure safety. This only adds to my lowered standards when approaching NYT articles over the years. They've long been abandoning being correct for being entertaining at a seemingly accelerating rate. Which is too bad considering plenty of people continue to hold them to a higher credibility over the typical clickbaiter sites. Considering they are a subscription service I'm not convinced this is justified.
- CivBase 8y ago
  > To be fair the article doesn't critique using 1990s (or earlier) hardware/software. I actually edited in an extra paragraph at the end of my original post because I wanted to at least credit the author for talking more about design issues than "processing power" in the body of the article. The title is very bad, though.
- huffmsa 8y ago
  The NYT fired basically their entire editorial staff in 2017. The accelerated degradation has been noticable.
  - DanielBMarkham 8y ago
    Everybody's talking about how we're going to pay for good content. How will newspapers operate in the future? We don't need to pay for content. The world is full of people who will write for nothing more than getting their name out there. What we need are good editors. The news media are firing the very people who could get them out of this mess they ended up in.
    - IronWolve 8y ago
      I pay for good content, lots of writers/producers use crowdfunding for a paycheck. Papers like NYTimes have slid down to clickbait titles and Buzzfeed type articles to get readers. The war for media control is corporate vs independent.
      - DanielBMarkham 8y ago
        I disagree. We've had independents, corporations, and mixed-mode content creation and delivery for many, many decades. What we used to have that we don't have any more is a forced-delay, quality-control, gatekeeping function. This was fulfilled by the slowness of manufacturing printed content and good editors. Turns out that this is a place where cutting out the middleman, the wisdom of crowds, and automation worked against our best interests.
  - Arnt 8y ago
    What? I googled, and found this: "The Times newsroom staff peaked in early 2008 at about 1,332 employees, [...] still employs about 1,300 in its newsroom." What are the correct numbers?
  - et2o 8y ago
    This is a good and comprehensive article. You and the grandparent comment are focusing on one minor component and using it to raise FUD about the rest of the article and indeed the entire NYT now. It's a lazy form of debate. HN used to be better than this.
- arethuza 8y ago
  I remember working in the early 90s on a training simulation of the reactor safety shutdown equipment for a UK AGR reactor - this device was pretty critical as it as responsible for sequencing the shutdown operations during a scram. It was designed in the 1960s using technology they were completely confident would work and keep working - so it was basically 1940s electro-mechanical technology: relays, worm gears etc. I don't think they really trusted electronics at that point, let alone computing devices for something so super critical. No idea how that is done now though!
- throw0101a 8y ago
  > To be fair the article doesn't critique using 1990s (or earlier) hardware/software. It's merely a clickbait headline. The "1990s home computers" power was simply an illustration of Boeing just trying to do the minimum work to move forward. Other examples that, while not bad per se, seem to indicate shortcuts: > If one of two sensors malfunctioned, the system could struggle to know which was right. Airbus addressed this potential problem on some of its planes by installing three or more such sensors. > Most new Boeing jets have electronic systems that take pilots through their preflight checklists, ensuring they don’t skip a step and potentially miss a malfunctioning part. On the Max, pilots still complete those checklists manually in a book. > A second electronic system found on other Boeing jets also alerts pilots to unusual or hazardous situations during flight and lays out recommended steps to resolve them. On 737s, a light typically indicates the problem and pilots have to flip through their paper manuals to find next steps. [...] Boeing decided against adding it to the Max because it could have prompted regulators to require new pilot training, according to two former Boeing employees involved in the decision. All of these things still get the job done, but they seem to indicate that Boeing simply wanted to do one last turn of the crank with the 737 design to compete with the A320neo, and may have rushed thing too much.
jschwartzi 8y ago
Yes, the 486 or 586 are far more powerful processors than what you would actually need for everything but the cockpit UI. You could probably get away with a Z80 or something similar. The flight control computers in our space probes frequently have a similar level of functionality simply because it's more than enough. The only thing a more modern processor buys you is more memory, a faster tick rate for your RTOS, potentially an MMU if it doesn't cause substantial timing skew, and high-precision floating point. In the case of an avionics system I would imagine the requirements for fixed-point arithmetic are very well-understood at this point, so floating point doesn't buy you a whole lot and potentially creates new and exciting problems. So I would say that 90's-era computers are plenty powerful for what they're used for.
- bluedino 8y ago
  The goddamn space shuttle ran on 386's, and that was an upgrade from 8086's
  - bdavis__ 8y ago
    i think the space shuttle used a couple of AP101's, which were 16 bit machines, with core memory, programmed in JOVIAL...
- kakwa_ 8y ago
  IIRC, the Airbus A320 uses several older CPUs like that, but also Motorola 68k CPUs (sometimes as a mean to have 2 functionally identical systems with pieces from different manufacturers and different implementations). I would not be surprised if they still use 68k in the A320 NEO. Between specification, implementation, validation and testing, the cycles in the aircraft industry can be quite long. 5 to 10 years usually. And it's also quite costly to validate a brand new system. That being said, given the safety requirement, it's for the better.
manojlds 8y ago
We sent people to the moon in 1969. We can definitely live with 90s computing power when it's enough.
crimsonalucard 8y ago
I think the author was referring to processors handling UI displays rather than a computer system that has only the processing power to only handle a blinking LED. For example if a UI display showed: "Warning: MCAS overriding pitch up. Disable?" Is better than a LED #234 and LED 412 are on. Consult manual for instructions. I would imagine that a lot of the fly by wire or control systems nowadays are either analog systems or very very simple digital systems (with extreme redundancy).
- jamesdgb 8y ago
  I understood the author the same way you did
- theclaw 8y ago
  It's a moot point anyway - the could not change the UI beyond converting physical dials to virtual ones, because that would have required retraining which they wanted to avoid. Even if they could have changed the UI they still needed to hide the existence of the MCAS system, again to avoid retraining, so it's doubtful such a clear an helpful error message would be allowable.
  - ssully 8y ago
    It isn't a moot point, it is literally the point of the article. The whole article is explaining the history of the 737 and explaining why they designed the max to avoid retraining. It mentions digital UIs, fly-by-wire, and other designs found in modern jets to paint the picture that the Max was successful in avoiding retraining, but that requirement restricted the engineers and led to a potentially compromised design.
- CivBase 8y ago
  You might be right. However, the author didn't show any evidence that the UI was limited by the available processing power. 1990s computers were very capable of displaying advanced graphic interfaces.
iSnow 8y ago
>Why exactly does this author think an airplane needs a lot of computing power? I think the way you criticize the article is about as faulty as the article itself. This piece is about detailing how the 737 is an outdated design that got patched and patched till it more or less broke. The processing power is just /one/ aspect and is given as a point of reference, like the 1968 photo. I am no expert in aviation, but I guess it's quite a safe guess that fly-by-wire systems need more processing power than mechanical/analog designs. And that a glass cockpit with a trouble database needs more than a blinking LED and a paper handbook. Nowhere in the article does the author claim the 737 MAX crashed because it had too weak CPUs. It is, however, one indication of an obsolete design - newer jets like the A320 or the 787 have more beefy processors for a reason. I know I worked in 1-dimensional transportation and their 486-based onboard-processors felt cramped, the software engineers had to hack around their limitations and they wished they could switch to something newer. I doubt 3-dimensional transportation is easier to implement.
- et2o 8y ago
  Thank you for taking the time to write this critique of the parent. You're totally correct.
- CivBase 8y ago
  > This piece is about detailing how the 737 is an outdated design that got patched and patched till it more or less broke. The processing power is just /one/ aspect and is given as a point of reference, like the 1968 photo. I address that in my post and I do not fault the author for pointing out design issues. My issue is with the implication that processing power is an issue. > Nowhere in the article does the author claim the 737 MAX crashed because it had too weak CPUs. It is, however, one indication of an obsolete design - newer jets like the A320 or the 787 have more beefy processors for a reason. Newer jets have faster processors because they're built on a newer system. If you're developing a new system, you might as well throw in some faster hardware. The cost of the hardware itself is nothing compared to the overhead development costs for avionics.
  - iSnow 8y ago
Bubbadoo 8y ago
Modern aviation is characterized by huge networks of sensors as most flight management systems are, in fact, compute driven and, since they are real time, higher compute power reduces latency. Clearly, this version of the Boeing 737 was produced with cost containment as the first priority.
- CivBase 8y ago
  Is there any evidence that latency is an issue?

dmix 8y ago

Wait until they finds out what software powers the nuclear weapons of 90%+ of the world, top-tier fighter jets, "critical" infrastructure like power (including nuclear), and countless financial/banking systems. To be fair the article doesn't critique using 1990s (or earlier) hardware/software. It's merely a clickbait headline. The only reference is dismissed by the very next sentence as insignificant [Boeing could be replaced with any ~~expert~~ somewhat-knowledgable person on the subject]: > The flight-control computers have roughly the processing power of 1990s home computers. A Boeing spokesman said the aircraft was designed with an appropriate level of technology to ensure safety. This only adds to my lowered standards when approaching NYT articles over the years. They've long been abandoning being correct for being entertaining at a seemingly accelerating rate. Which is too bad considering plenty of people continue to hold them to a higher credibility over the typical clickbaiter sites. Considering they are a subscription service I'm not convinced this is justified.

CivBase 8y ago
> To be fair the article doesn't critique using 1990s (or earlier) hardware/software. I actually edited in an extra paragraph at the end of my original post because I wanted to at least credit the author for talking more about design issues than "processing power" in the body of the article. The title is very bad, though.
huffmsa 8y ago
The NYT fired basically their entire editorial staff in 2017. The accelerated degradation has been noticable.
- DanielBMarkham 8y ago
  Everybody's talking about how we're going to pay for good content. How will newspapers operate in the future? We don't need to pay for content. The world is full of people who will write for nothing more than getting their name out there. What we need are good editors. The news media are firing the very people who could get them out of this mess they ended up in.
  - IronWolve 8y ago
    I pay for good content, lots of writers/producers use crowdfunding for a paycheck. Papers like NYTimes have slid down to clickbait titles and Buzzfeed type articles to get readers. The war for media control is corporate vs independent.
    - DanielBMarkham 8y ago
      I disagree. We've had independents, corporations, and mixed-mode content creation and delivery for many, many decades. What we used to have that we don't have any more is a forced-delay, quality-control, gatekeeping function. This was fulfilled by the slowness of manufacturing printed content and good editors. Turns out that this is a place where cutting out the middleman, the wisdom of crowds, and automation worked against our best interests.
- Arnt 8y ago
  What? I googled, and found this: "The Times newsroom staff peaked in early 2008 at about 1,332 employees, [...] still employs about 1,300 in its newsroom." What are the correct numbers?
- et2o 8y ago
  This is a good and comprehensive article. You and the grandparent comment are focusing on one minor component and using it to raise FUD about the rest of the article and indeed the entire NYT now. It's a lazy form of debate. HN used to be better than this.
arethuza 8y ago
I remember working in the early 90s on a training simulation of the reactor safety shutdown equipment for a UK AGR reactor - this device was pretty critical as it as responsible for sequencing the shutdown operations during a scram. It was designed in the 1960s using technology they were completely confident would work and keep working - so it was basically 1940s electro-mechanical technology: relays, worm gears etc. I don't think they really trusted electronics at that point, let alone computing devices for something so super critical. No idea how that is done now though!
throw0101a 8y ago
> To be fair the article doesn't critique using 1990s (or earlier) hardware/software. It's merely a clickbait headline. The "1990s home computers" power was simply an illustration of Boeing just trying to do the minimum work to move forward. Other examples that, while not bad per se, seem to indicate shortcuts: > If one of two sensors malfunctioned, the system could struggle to know which was right. Airbus addressed this potential problem on some of its planes by installing three or more such sensors. > Most new Boeing jets have electronic systems that take pilots through their preflight checklists, ensuring they don’t skip a step and potentially miss a malfunctioning part. On the Max, pilots still complete those checklists manually in a book. > A second electronic system found on other Boeing jets also alerts pilots to unusual or hazardous situations during flight and lays out recommended steps to resolve them. On 737s, a light typically indicates the problem and pilots have to flip through their paper manuals to find next steps. [...] Boeing decided against adding it to the Max because it could have prompted regulators to require new pilot training, according to two former Boeing employees involved in the decision. All of these things still get the job done, but they seem to indicate that Boeing simply wanted to do one last turn of the crank with the 737 design to compete with the A320neo, and may have rushed thing too much.

jschwartzi 8y ago

Yes, the 486 or 586 are far more powerful processors than what you would actually need for everything but the cockpit UI. You could probably get away with a Z80 or something similar. The flight control computers in our space probes frequently have a similar level of functionality simply because it's more than enough. The only thing a more modern processor buys you is more memory, a faster tick rate for your RTOS, potentially an MMU if it doesn't cause substantial timing skew, and high-precision floating point. In the case of an avionics system I would imagine the requirements for fixed-point arithmetic are very well-understood at this point, so floating point doesn't buy you a whole lot and potentially creates new and exciting problems. So I would say that 90's-era computers are plenty powerful for what they're used for.

bluedino 8y ago
The goddamn space shuttle ran on 386's, and that was an upgrade from 8086's
- bdavis__ 8y ago
  i think the space shuttle used a couple of AP101's, which were 16 bit machines, with core memory, programmed in JOVIAL...
kakwa_ 8y ago
IIRC, the Airbus A320 uses several older CPUs like that, but also Motorola 68k CPUs (sometimes as a mean to have 2 functionally identical systems with pieces from different manufacturers and different implementations). I would not be surprised if they still use 68k in the A320 NEO. Between specification, implementation, validation and testing, the cycles in the aircraft industry can be quite long. 5 to 10 years usually. And it's also quite costly to validate a brand new system. That being said, given the safety requirement, it's for the better.

manojlds 8y ago

We sent people to the moon in 1969. We can definitely live with 90s computing power when it's enough.

crimsonalucard 8y ago

I think the author was referring to processors handling UI displays rather than a computer system that has only the processing power to only handle a blinking LED. For example if a UI display showed: "Warning: MCAS overriding pitch up. Disable?" Is better than a LED #234 and LED 412 are on. Consult manual for instructions. I would imagine that a lot of the fly by wire or control systems nowadays are either analog systems or very very simple digital systems (with extreme redundancy).

jamesdgb 8y ago
I understood the author the same way you did
theclaw 8y ago
It's a moot point anyway - the could not change the UI beyond converting physical dials to virtual ones, because that would have required retraining which they wanted to avoid. Even if they could have changed the UI they still needed to hide the existence of the MCAS system, again to avoid retraining, so it's doubtful such a clear an helpful error message would be allowable.
- ssully 8y ago
  It isn't a moot point, it is literally the point of the article. The whole article is explaining the history of the 737 and explaining why they designed the max to avoid retraining. It mentions digital UIs, fly-by-wire, and other designs found in modern jets to paint the picture that the Max was successful in avoiding retraining, but that requirement restricted the engineers and led to a potentially compromised design.
CivBase 8y ago
You might be right. However, the author didn't show any evidence that the UI was limited by the available processing power. 1990s computers were very capable of displaying advanced graphic interfaces.

iSnow 8y ago

>Why exactly does this author think an airplane needs a lot of computing power? I think the way you criticize the article is about as faulty as the article itself. This piece is about detailing how the 737 is an outdated design that got patched and patched till it more or less broke. The processing power is just /one/ aspect and is given as a point of reference, like the 1968 photo. I am no expert in aviation, but I guess it's quite a safe guess that fly-by-wire systems need more processing power than mechanical/analog designs. And that a glass cockpit with a trouble database needs more than a blinking LED and a paper handbook. Nowhere in the article does the author claim the 737 MAX crashed because it had too weak CPUs. It is, however, one indication of an obsolete design - newer jets like the A320 or the 787 have more beefy processors for a reason. I know I worked in 1-dimensional transportation and their 486-based onboard-processors felt cramped, the software engineers had to hack around their limitations and they wished they could switch to something newer. I doubt 3-dimensional transportation is easier to implement.

et2o 8y ago
Thank you for taking the time to write this critique of the parent. You're totally correct.
CivBase 8y ago
> This piece is about detailing how the 737 is an outdated design that got patched and patched till it more or less broke. The processing power is just /one/ aspect and is given as a point of reference, like the 1968 photo. I address that in my post and I do not fault the author for pointing out design issues. My issue is with the implication that processing power is an issue. > Nowhere in the article does the author claim the 737 MAX crashed because it had too weak CPUs. It is, however, one indication of an obsolete design - newer jets like the A320 or the 787 have more beefy processors for a reason. Newer jets have faster processors because they're built on a newer system. If you're developing a new system, you might as well throw in some faster hardware. The cost of the hardware itself is nothing compared to the overhead development costs for avionics.
- iSnow 8y ago

Bubbadoo 8y ago

Modern aviation is characterized by huge networks of sensors as most flight management systems are, in fact, compute driven and, since they are real time, higher compute power reduces latency. Clearly, this version of the Boeing 737 was produced with cost containment as the first priority.

CivBase 8y ago
Is there any evidence that latency is an issue?