| 1 |
> > So, we're considering some system that (we think so here) can |
| 2 |
> > guarantee meeting its deadlines with a given (by the operator) probability. |
| 3 |
|
| 4 |
> You cannot prove determism with statistics. |
| 5 |
> Determinism can only be proven be defining, discovering, |
| 6 |
> characterizing and testing each and ever possible state AND |
| 7 |
> transition. |
| 8 |
|
| 9 |
What would you say if our considered system, say, can mathematically |
| 10 |
achieve this (please see below, next statement) |
| 11 |
|
| 12 |
|
| 13 |
> > We imply here that true HRT is impossible in reality and any single |
| 14 |
> > system that exists in the word meets hard real-time requirements with |
| 15 |
> > some probability. If it does not – HRT speaks of system failure. |
| 16 |
|
| 17 |
> Ah, how you are 'getting it'. YES, YES, YES, I absolutely agree |
| 18 |
> with this statement! |
| 19 |
|
| 20 |
That's a first step we must to achieve in our discussion :). |
| 21 |
I'm aware that you think so. I'm also thinking so. But we were wrongly |
| 22 |
messed up with fault tolerance before… We know now, that |
| 23 |
[any HRT system CAN miss its deadlines when in a faulty state] – |
| 24 |
that's a statement. |
| 25 |
|
| 26 |
Then, lets take a look at our considered system as of blackbox to the |
| 27 |
external world. The world knows nothing of its internals, state |
| 28 |
machines, transitions, and languages. The world is just looking at the |
| 29 |
system in terms of request-responses and timing constrains, it is |
| 30 |
expecting. |
| 31 |
|
| 32 |
Say, the world thinks of the system must to work for it with some |
| 33 |
predictable reliability. Please note, missed deadlines are just faults |
| 34 |
here. |
| 35 |
|
| 36 |
If we can mathematically prove the system is reliable not worse than |
| 37 |
expected reliability, may we think it is HRT? |
| 38 |
|
| 39 |
-- |
| 40 |
gentoo-embedded@g.o mailing list |
Archives