1 |
momentics wrote: |
2 |
> so, from the above, your point is that HRT system should have no |
3 |
> laxity on its deadlines? |
4 |
|
5 |
I think the very nature of the definition is that a HRT system |
6 |
meets or exceeds timing constraints. If you say a HRT system |
7 |
is 150ns +/- 10%, you are really saying your absolute |
8 |
timing deadline is 165ns, to as many significant digits |
9 |
as you like. |
10 |
|
11 |
The mere fact you are asking/questioning with the term |
12 |
'laxity' means it's not a HRT. A HRT has to be defined |
13 |
with some sort of absolute timing constraints, otherwise |
14 |
it breaks down to a SRT system with arbitrary constraints. |
15 |
|
16 |
> ie the granularity of measures lim(granularity)-> zero |
17 |
|
18 |
Well, this is a skewed representation. I like to think of |
19 |
it more like a logic based system. It either meets or exceeds |
20 |
timing constraints, or it does not. |
21 |
|
22 |
> like a mathematical abstraction (little's law) that it is impossible |
23 |
> to catch a tortoise up… |
24 |
|
25 |
Um, that story has lots of parameters that are not characterized |
26 |
as merely timing or deadlines, like the unpredictable/irrational |
27 |
behavior of the fox, Who cold have merely trotted into compliance |
28 |
(beaten the tortoise) and still have enjoyed frivolity along the |
29 |
journey. And yes that race did not have any specific timing |
30 |
constraint it was 'best effort' as we know there was not a specific |
31 |
allotted amount of time to finish, just who crossed the finish line |
32 |
first. That said, this story does mimic the arguments about hard/soft |
33 |
real time, as they are sheer folly..... |
34 |
|
35 |
It's all about latency, reliability(demonstrated via reproducible, |
36 |
timing constraint results), and on rare occasions, deterministic sub |
37 |
systems. A system that never fails is fault tolerant, until it |
38 |
actually fails. Just look at the telecom industry. Five nines of |
39 |
uptime (less than 5 minutes of down time per year on a phone switch is |
40 |
what the industry established, as practical matter of fault tolerance |
41 |
on a HA (high availability) system. It's not mathematically correct, |
42 |
but, it's good for business, and a practical trade off. |
43 |
Technical folks, particularly computer scientists, should |
44 |
fundamentally understand where science ends and where pop-science |
45 |
begins, in my opinion. |
46 |
|
47 |
|
48 |
James |
49 |
|
50 |
-- |
51 |
gentoo-embedded@g.o mailing list |