G'day Karl,
Let's say I have three clients (users of my MCE application): A has 4 lcores; B has 8 lcores; C has 16 lcores.
Let's further say that each client only wants to use half of their processing power for my application, so:
A wants to use 2 lcores; B wants to use 4 lcores; C wants to use 8 lcores.
If I simply use an "int param", as you suggest, they'll get the same value for max_workers.
For instance, if I used max_workers => 6:
A would be using 150%; B would be using 75%; C would be using 37.5%.
(Compare what happened, in the first half of my OP, where I didn't cap the value and, at 200%, my computer
with 12 lcores got 24 workers.)
Furthermore, clients' requirements may change depending on a number of factors.
Perhaps start a big job on a Friday afternoon, using all available processing power, and let it run over the weekend;
in another scenario, run jobs mid-week, during a busy part of the day, using only a fraction of available processing power.
Both of my solutions calculate the percentages correctly.
What I didn't know was if there was a better solution;
and, if not, which of those two solutions fellow monks would consider to be better
(keeping in mind, of course, that better might vary under different conditions).
| [reply]
[d/l]
[select] |
| [reply] |