[SystemSafety] Difference between software reliability and astrology

[Prof. Dr. Peter Bernard Ladkin]

On 2024-08-21 10:23 , Paul Sherwood wrote:
> On 2024-08-20 21:13, Prof. Dr. Peter Bernard Ladkin wrote:
>
> Hmmm. As I understand it [the memoryless distributions] are models, chosen by the modeller(s)? In 
> some sense these distributions actually describe what is observed in some classes of software 
> behaviour, but your papers eloquently describe various limitations.

You can leave out the words "in some sense". If your software behaviour fulfils the "memoryless" 
requirement, the failure behaviour of your on-demand software is a Bernoulli process and the failure 
behaviour of your continuously-operating software is a Poisson process. If your software behaviour 
doesn't fulfil the "memoryless" requirement, then with on-demand functions you are largely out of 
luck in finding a workable stochastic process which it fits, and with continously-operating 
functions you can maybe show that parts of it fulfil the conditions of another renewal process and 
use the limit theorem to argue that the overall is thus Poisson.

>
> It's obvious that some software behaviour doesn't fulfil the property, so if we hope to make 
> progress there we would need to identify/devise/choose different models?

If you can find one which can be worked with. The fact that so far no one has done so might bring 
you up short. Mathematical statistics is a competitive field.

>>> "We conclude that establishing the reliability of RTOS practically using the Bernoulli/Poisson 
>>> mathematics in this manner looks close to infeasible. Yet Annex D currently states in its second 
>>> sentence “This approach is considered particularly appropriate as part of the qualification of 
>>> operating systems, [etc.]” !
>>>
>>> It seems to me that for complex software in general, we'll need something better?
>>
>> Better? Like what?
>
> https://mathworld.wolfram.com/BayesianAnalysis.html perhaps?

That is a category mistake. Bayesian analysis is a form of statistical inference, it's not a 
collection of different stochastic processes.

PBL

Prof. Dr. Peter Bernard Ladkin
Causalis Limited/Causalis IngenieurGmbH, Bielefeld, Germany
Tel: +49 (0)521 3 29 31 00