Re: [PATCH 00/13] Reconcile NUMA balancing decisions with the load balancer v3

[Mel Gorman]

On Mon, Feb 17, 2020 at 03:52:16PM +0100, Peter Zijlstra wrote:
> On Mon, Feb 17, 2020 at 02:49:11PM +0100, Vincent Guittot wrote:
> > > Patches 4-5 are Vincent's and use very similar code patterns and logic
> > > between NUMA and load balancer. Patch 6 is a fix to Vincent's work that
> > > is necessary to avoid serious imbalances being introduced by the NUMA
> > 
> > Yes the test added in load_too_imbalanced() by patch 5 doesn't seem to
> > be a good choice.
> > I haven't remove it as it was done by your patch 6 but it might worth
> > removing it directly if a new version is needed
> 
> Aside of that, Vincent's patches look good to me.

Fully agreed, I think it's now much easier to understand the two balancers
when put side by side in addition to getting some performance gains.
Even if a regression is found, I think it'll be due to a workload seeing
an advantage when NUMA balancer constantly overrides the load balancer.
If that happens, the imbalance can be governed by adjust_numa_balance so
that the two balancers avoid fighting each other again.

-- 
Mel Gorman
SUSE Labs