Re: No, really, stop trying to delete slab until you've finished making slub perform as well

[Christoph Lameter <cl@linux-foundation.org>]

KOSAKI Motohiro wrote:

> new patch is here.
> 
> Index: b/mm/slub.c
> ===================================================================
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -1326,9 +1326,11 @@ static struct page *get_any_partial(stru
>          * expensive if we do it every time we are trying to find a slab
>          * with available objects.
>          */
> +#if 0
>         if (!s->remote_node_defrag_ratio ||
>                         get_cycles() % 1024 > s->remote_node_defrag_ratio)
>                 return NULL;
> +#endif
> 
>         zonelist = node_zonelist(slab_node(current->mempolicy), flags);
>         for_each_zone_zonelist(zone, z, zonelist, high_zoneidx) {

Hmmm.... So always take from partial lists works? That is the same effect that
the setting of the remote_defrag_ratio to 100 should have had (its multiplied
by 10 when storing it).

So its a NUMA only phenomenon. How is performance affected?