Nick Piggin wrote:
> If cc1 is doing a lot of waiting on IO, I fail to see how it should be
> called a CPU hog. OK I'll stop being difficult! I understand the problem
> is that its behaviour suddenly changes from IO bound to CPU hog, right?
> Then it seems like the scheduler's problem is that it doesn't adapt
> quickly enough to this change.
>
> What you are doing is restricting some range so it can adapt more quickly
> right? So you still have the problem in the cases where you are not
> restricting this range.
For this, I reiterate my suggestion to intentionally over-shoot the
mark. If you do it right, a process will run an inappropriate length of
time only every other time slice until the oscillation dies down.
Let me give you an example. Let's say you have a process which is being
interactive, and then suddenly becomes a CPU hog.
In the case as it is (assumptions here), what happens is that the
priority is reduced by some amount until it reaches a level appropriate
for the new behavior.
I get the impression that lower numbers mean higher priority, so here goes:
- The process starts out with a priority of 10 (this may mean something
that I don't know about... just follow along).
- It becomes a CPU hog sufficient to make it NEED to be at a priority of 30.
- Over some number of time slices, the priority is changed something
like this: 10, 20, 25, 27, 28, 29, 30.
Here's my alternative suggestion -- if 10 is pure interactive and 30 is
CPU hog, and you see some change in behavior, before, you would go half
way. Now, instead, go one-and-a-half way.
- Over some number of time slices, the priority is changed like this:
10, 40, 25, 32, 27, 31, 28, 30
Let's say that you only get one time slice which is CPU hog, but others
are not, for the first case, you'd get something like this:
10, 20, 15, 12, 11, 10
For the second case, you'd get this:
10, 40, -5, 17, 7, 11, 10
Something like that. So instead of getting tricked and having to
return, it over shoots but makes up for it the next time the process is run.
This is a very incomplete thought and may be pure garbage, so please
forgive me if I'm being an idiot. :)
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Aug 07 2003 - 22:00:35 EST