Re: [PATCH 00/10] steal tasks to improve CPU utilization

[Peter Zijlstra]

On Mon, Oct 22, 2018 at 03:07:10PM -0400, Steven Sistare wrote:
> On 10/22/2018 1:04 PM, Peter Zijlstra wrote:
> > On Mon, Oct 22, 2018 at 07:59:31AM -0700, Steve Sistare wrote:
> >> When a CPU has no more CFS tasks to run, and idle_balance() fails to
> >> find a task, then attempt to steal a task from an overloaded CPU in the
> >> same LLC. Maintain and use a bitmap of overloaded CPUs to efficiently
> >> identify candidates.  To minimize search time, steal the first migratable
> >> task that is found when the bitmap is traversed.  For fairness, search
> >> for migratable tasks on an overloaded CPU in order of next to run.
> >>
> >> This simple stealing yields a higher CPU utilization than idle_balance()
> >> alone, because the search is cheap, so it may be called every time the CPU
> >> is about to go idle.  idle_balance() does more work because it searches
> >> widely for the busiest queue, so to limit its CPU consumption, it declines
> >> to search if the system is too busy.  Simple stealing does not offload the
> >> globally busiest queue, but it is much better than running nothing at all.
> > 
> > Why I don't dislike the idea; I feel it is unfortunate to have two
> > different mechanisms to do effectively the same thing.
> > 
> > Can't we improve idle_balance() instead of building this parallel
> > functionality?
> 
> We could delete idle_balance() and use stealing exclusively for handling
> new idle.  For each sd level, stealing would look for an overloaded CPU
> in the overloaded bitmap(s) that overlap that level.  I played with that
> a little but it is not ready for prime time, and I did not want to hold
> the patch series for it.  Also, I would like folks to get some production
> experience with stealing on a variety of architectures before considering
> a radical step like replacing idle_balance().

Fair enough. And yes, it might make sense to fully replace the current
newidle balance with something along these lines.

> We could remove the core and socket levels from idle_balance() and let
> stealing handle those levels.  I think that makes sense after stealing
> performance is validated on more architectures, but we would still have
> two different mechanisms.

Yes, this would be a fairly simple change and make sense until we have a
full replacement.

> We could merge the stealing code into the idle_balance() code to get a
> union of the two, but IMO that would be less readable.

Agreed; I don't think that'll be pretty.