Re: [patch] CFS scheduler, v3

[Peter Williams <pwil3058@bigpond.net.au>]

Peter Williams wrote:
> Ingo Molnar wrote:
> >  - bugfix: use constant offset factor for nice levels instead of    
> > sched_granularity_ns. Thus nice levels work even if someone sets    
> > sched_granularity_ns to 0. NOTE: nice support is still naive, i'll    
> > address the many nice level related suggestions in -v4.
>
> I have a suggestion I'd like to make that addresses both nice and 
> fairness at the same time.  As I understand the basic principle behind 
> this scheduler it to work out a time by which a task should make it onto 
> the CPU and then place it into an ordered list (based on this value) of 
> tasks waiting for the CPU.  I think that this is a great idea and my 
> suggestion is with regard to a method for working out this time that 
> takes into account both fairness and nice.
>
> First suppose we have the following metrics available in addition to 
> what's already provided.
>
> rq->avg_weight_load /* a running average of the weighted load on the CPU */
> p->avg_cpu_per_cycle /* the average time in nsecs that p spends on the 
> CPU each scheduling cycle */
>
> where a scheduling cycle for a task starts when it is placed on the 
> queue after waking or being preempted and ends when it is taken off the 
> CPU either voluntarily or after being preempted.  So 
> p->avg_cpu_per_cycle is just the average amount of time p spends on the 
> CPU each time it gets on to the CPU.  Sorry for the long explanation 
> here but I just wanted to make sure there was no chance that "scheduling 
> cycle" would be construed as some mechanism being imposed on the 
> scheduler.)
>
> We can then define:
>
> effective_weighted_load = max(rq->raw_weighted_load, rq->avg_weighted_load)
>
> If p is just waking (i.e. it's not on the queue and its load_weight is 
> not included in rq->raw_weighted_load) and we need to queue it, we say 
> that the maximum time (in all fairness) that p should have to wait to 
> get onto the CPU is:
>
> expected_wait = p->avg_cpu_per_cycle * effective_weighted_load / 
> p->load_weight

I just realized that this is wrong for the case where p is being woken. 
 In that case the length of the last sleep should be subtracted from 
the above value and if the result is negative pre-empt straight away. 
So expected_wait becomes:

expected_wait = p->avg_cpu_per_cycle * effective_weighted_load / 
p->load_weight - p->length_of_last_sleep

As the length of the last sleep will be being calculated during the wake 
process there's no real need for it to be a task field and a local 
variable could be used instead.

For a task being requeued during pre-emption the equation would be:

expected_wait = time_just_spent_on_the_cpu * effective_weighted_load / 
p->load_weight

as there was zero sleep since last time on CPU.  Using the actual time 
on the CPU so far this time (instead of the average) will compensate the 
task for being pre-empted.

> Calculating p->avg_cpu_per_cycle costs one add, one multiply and one 
> shift right per scheduling cycle of the task.  An additional cost is 
> that you need a shift right to get the nanosecond value from value 
> stored in the task struct. (i.e. the above code is simplified to give 
> the general idea).  The average would be number of cycles based rather 
> than time based and (happily) this simplifies the calculations.

If you don't like using the average CPU time per cycle, you could just 
use the length of time on the CPU for the last time the task was on the 
CPU.  This would simplify things a lot.

I've been thinking about the "jerkiness" (or lack of "smoothness") that 
I said would occur if smoothed averages weren't used and I've realised 
that what I was talking about was observed jerkiness in the dynamic 
priorities of the tasks.  As this scheduler has dispensed with dynamic 
priorities (as far as I can see) the jerkiness probably won't be apparent.

BTW Given that I'm right and dynamic priorities have been dispensed with 
what do you intend exporting (in their place) to user space for display 
by top and similar?

Peter
--
Peter Williams                                   pwil3058@bigpond.net.au

"Learning, n. The kind of ignorance distinguishing the studious."
 -- Ambrose Bierce
-
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/