4 ms·

I wonder the same for large (especially international) companies, government bureaucracies or other organizations. With thousands of employees and branches in d

I wonder the same for large (especially international) companies, government bureaucracies or other organizations. With thousands of employees and branches in different locations, it must be impossible to get a full understanding of the state of the system.

auto 5y ago
Same, makes you wonder how much slack/tolerance is built into these systems and organizations, and whether that comes in the form of manpower, process, etc. and whether or not it’s planned for, or just naturally evolves in large enough instances.
jasonwatkinspdx 5y ago
This is what Systems Engineers do. There's no magic bullet, but the basic process is lots and lots of documents. Much of what NASA does is published openly so you can get a feel for what some of these document types are like. The wiki page gives a decent summary. Building a model for the system is indeed a big part of the challenge. These days it's increasingly done as a fully detailed software simulation. The big CAD packages have specific functionality for this now. For example SpaceX runs on Siemens NX, and you can wade through their marketing speak to get some idea of how it works. CATIA is also popular with aerospace companies, and Autocad's products with architecture.
- sydthrowaway 5y ago
  Ugh, I can't deal with Systems Engineering. It's a whole lot of gobbledygook before getting to the crux of the problem. A software engineering analogue would be FizzBuzz Enterprise Edition https://github.com/EnterpriseQualityCoding/FizzBuzzEnterpriseEdition https://github.com/EnterpriseQualityCoding/FizzBuzzEnterpris...
  - spaetzleesser 5y ago
    I don’t like it either but in regulated industries like space, medical or aerospace it seems the only way to manage things in a controlled way.
  - jasonwatkinspdx 5y ago
    I can understand not liking how bureaucratically onerous it is, but we don't really have a choice. It's imperfect but it's the only way we've found to manage highly complex large scale engineering projects. You certainly aren't going to build something like JWT with a just YOLO agile process unless you're literally willing to fail like 50 times before you get one right (aka the USSR approach to rockets). The upside is automation is making it go smoother and be less burdensome, because simulated testing truly is high enough fidelity now it allows a bit more virtual trial and error in the process.
    - bumby 5y ago
      >The upside is automation is making it go smoother and be less burdensome, because simulated testing truly is high enough fidelity I agree to an extent, but there's also a risk of simulation breeding a false sense of security even when simulations are conducted well. The investigation of the CST-100 "anomalous" test flight had 21 findings related to software simulations and testing, some related to lack of fidelity. Not that fidelity wasn't possible, but it has some overtones of the Ariane 5 software issue in that there was a lack of integration testing within the different software components.
      - spaetzleesser 5y ago
        The good thing about simulations is that once you have figured out a problem it won’t make the same mistake again. That’s the strength of computing
        bumby 5y ago
        I agree, but the problem with this case was that simulations led to complacency about not thinking they needed integration tests. That's much rarely in the hardware domain. It's not really meant to be a critique of simulations but rather how we use them.
- spaetzleesser 5y ago
  “ There's no magic bullet, but the basic process is lots and lots of documents.” That’s probably it. My company has systems engineers and even in relatively small projects you often have inconsistent and incomplete requirements. It’s one thing to design a piece of hardware but you also need to track how logistics and other factors have an impact on the system. So you need a lot of people who constantly check these changes.

auto 5y ago

Same, makes you wonder how much slack/tolerance is built into these systems and organizations, and whether that comes in the form of manpower, process, etc. and whether or not it’s planned for, or just naturally evolves in large enough instances.

jasonwatkinspdx 5y ago

This is what Systems Engineers do. There's no magic bullet, but the basic process is lots and lots of documents. Much of what NASA does is published openly so you can get a feel for what some of these document types are like. The wiki page gives a decent summary. Building a model for the system is indeed a big part of the challenge. These days it's increasingly done as a fully detailed software simulation. The big CAD packages have specific functionality for this now. For example SpaceX runs on Siemens NX, and you can wade through their marketing speak to get some idea of how it works. CATIA is also popular with aerospace companies, and Autocad's products with architecture.

sydthrowaway 5y ago
Ugh, I can't deal with Systems Engineering. It's a whole lot of gobbledygook before getting to the crux of the problem. A software engineering analogue would be FizzBuzz Enterprise Edition https://github.com/EnterpriseQualityCoding/FizzBuzzEnterpriseEdition https://github.com/EnterpriseQualityCoding/FizzBuzzEnterpris...
- spaetzleesser 5y ago
  I don’t like it either but in regulated industries like space, medical or aerospace it seems the only way to manage things in a controlled way.
- jasonwatkinspdx 5y ago
  I can understand not liking how bureaucratically onerous it is, but we don't really have a choice. It's imperfect but it's the only way we've found to manage highly complex large scale engineering projects. You certainly aren't going to build something like JWT with a just YOLO agile process unless you're literally willing to fail like 50 times before you get one right (aka the USSR approach to rockets). The upside is automation is making it go smoother and be less burdensome, because simulated testing truly is high enough fidelity now it allows a bit more virtual trial and error in the process.
  - bumby 5y ago
    >The upside is automation is making it go smoother and be less burdensome, because simulated testing truly is high enough fidelity I agree to an extent, but there's also a risk of simulation breeding a false sense of security even when simulations are conducted well. The investigation of the CST-100 "anomalous" test flight had 21 findings related to software simulations and testing, some related to lack of fidelity. Not that fidelity wasn't possible, but it has some overtones of the Ariane 5 software issue in that there was a lack of integration testing within the different software components.
    - spaetzleesser 5y ago
      The good thing about simulations is that once you have figured out a problem it won’t make the same mistake again. That’s the strength of computing
      - bumby 5y ago
        I agree, but the problem with this case was that simulations led to complacency about not thinking they needed integration tests. That's much rarely in the hardware domain. It's not really meant to be a critique of simulations but rather how we use them.
spaetzleesser 5y ago
“ There's no magic bullet, but the basic process is lots and lots of documents.” That’s probably it. My company has systems engineers and even in relatively small projects you often have inconsistent and incomplete requirements. It’s one thing to design a piece of hardware but you also need to track how logistics and other factors have an impact on the system. So you need a lot of people who constantly check these changes.