Re: [PATCH 2/3] sched/numa: Allow a floating imbalance between NUMA nodes

[Mel Gorman]

On Tue, Nov 17, 2020 at 03:16:03PM +0100, Peter Zijlstra wrote:
> On Tue, Nov 17, 2020 at 01:42:21PM +0000, Mel Gorman wrote:
> > This patch revisits the possibility that NUMA nodes can be imbalanced
> > until 25% of the CPUs are occupied. The reasoning behind 25% is somewhat
> > superficial -- it's half the cores when HT is enabled.  At higher
> > utilisations, balancing should continue as normal and keep things even
> > until scheduler domains are fully busy or over utilised.
> 
> Do we want to make that shift depend on the actual SMT factor?

I considered it but decided against it. I wanted the balance point to
be somewhere below SMT because select_idle_sibling tries to avoid SMT
sharing so I didn't want a point where SMT sharing caused more problems
than memory locality. However, I worried that picking a different imbalance
point depending on SMT would be surprising to some.

-- 
Mel Gorman
SUSE Labs