Re: [RFC] A new CPU load metric for power-efficient scheduler: CPU ConCurrency

[Peter Zijlstra <peterz@infradead.org>]

On Fri, Apr 25, 2014 at 10:00:02AM +0200, Vincent Guittot wrote:
> On 24 April 2014 21:30, Yuyang Du <yuyang.du@intel.com> wrote:
> > Hi Ingo, PeterZ, and others,
> >
> > The current scheduler's load balancing is completely work-conserving. In some
> > workload, generally low CPU utilization but immersed with CPU bursts of
> > transient tasks, migrating task to engage all available CPUs for
> > work-conserving can lead to significant overhead: cache locality loss,
> > idle/active HW state transitional latency and power, shallower idle state,
> > etc, which are both power and performance inefficient especially for today's
> > low power processors in mobile.
> >
> > This RFC introduces a sense of idleness-conserving into work-conserving (by
> > all means, we really don't want to be overwhelming in only one way). But to
> > what extent the idleness-conserving should be, bearing in mind that we don't
> > want to sacrifice performance? We first need a load/idleness indicator to that
> > end.
> >
> > Thanks to CFS's "model an ideal, precise multi-tasking CPU", tasks can be seen
> > as concurrently running (the tasks in the runqueue). So it is natural to use
> > task concurrency as load indicator. Having said that, we do two things:
> >
> > 1)      Divide continuous time into periods of time, and average task concurrency
> > in period, for tolerating the transient bursts:
> > a = sum(concurrency * time) / period
> > 2)      Exponentially decay past periods, and synthesize them all, for hysteresis
> > to load drops or resilience to load rises (let f be decaying factor, and a_x
> > the xth period average since period 0):
> > s = a_n + f^1 * a_n-1 + f^2 * a_n-2 +, .....,+ f^(n-1) * a_1 + f^n * a_0
> 
> In the original version of entity load tracking patchset, there was a
> usage_avg_sum field that was counting the time the task was really
> running on the CPU. By combining this (disappeared ) field with the
> runnable_avg_sum, you should have similar concurrency value but with
> the current load tracking mechanism (instead of creating new one).

I'm not entire sure understood what was proposed, but I suspect its very
close to what I told you to do with the capacity muck. Use avg
utilization instead of 1 active task per core.

And yes, the current load tracking should be pretty close.

We just need to come up another way of doing SMT again, bloody
inconvenient SMT.