Re: [RFC 0/2] An attempt to improve SLUB on NUMA / under memory pressure

[Hyeonggon Yoo]

On Thu, Aug 10, 2023 at 7:56 PM Jay Patel <jaypatel@xxxxxxxxxxxxx> wrote:
>
> On Mon, 2023-07-24 at 04:09 +0900, Hyeonggon Yoo wrote:
> > Hello folks,
> >
> > This series is motivated by kernel test bot report [1] on Jay's patch
> > that modifies slab order. While the patch was not merged and not in
> > the
> > final form, I think it was a good lesson that changing slab order has
> > more
> > impacts on performance than we expected.
> >
> > While inspecting the report, I found some potential points to improve
> > SLUB. [2] It's _potential_ because it shows no improvements on
> > hackbench.
> > but I believe more realistic workloads would benefit from this. Due
> > to
> > lack of resources and lack of my understanding of *realistic*
> > workloads,
> > I am asking you to help evaluating this together.
>
> Hi Hyeonggon,
> I tried hackbench test on Powerpc machine with 16 cpus but
> got ~32% of Regression with patch.

Thank you so much for measuring this! That's very helpful.
It's interesting because on an AMD machine with 2 NUMA nodes there was
not much difference.

Does it have more than one socket?

Could you confirm if the offending patch is patch 1 or 2?
If the offending one is patch 2, can you please check how large is L3
cache miss rate
during hackbench?

> Results as
>
> +-------+----+---------+------------+------------+
> |       |    | Normal  | With Patch |            |
> +-------+----+---------+------------+------------+
> | Amean | 1  | 1.3700  | 2.0353     | ( -32.69%) |
> | Amean | 4  | 5.1663  | 7.6563     | (- 32.52%) |
> | Amean | 7  | 8.9180  | 13.3353    | ( -33.13%) |
> | Amean | 12 | 15.4290 | 23.0757    | ( -33.14%) |
> | Amean | 21 | 27.3333 | 40.7823    | ( -32.98%) |
> | Amean | 30 | 38.7677 | 58.5300    | ( -33.76%) |
> | Amean | 48 | 62.2987 | 92.9850    | ( -33.00%) |
> | Amean | 64 | 82.8993 | 123.4717   | ( -32.86%) |
> +-------+----+---------+------------+------------+
>
> Thanks
> Jay Patel
> >
> > It only consists of two patches. Patch #1 addresses inaccuracy in
> > SLUB's heuristic, which can negatively affect workloads' performance
> > when large folios are not available from buddy.
> >
> > Patch #2 changes SLUB's behavior when there are no slabs available on
> > the
> > local node's partial slab list, increasing NUMA locality when there
> > are
> > available memory (without reclamation) on the local node from buddy.
> >
> > This is early state, but I think it's a good enough to start
> > discussion.
> > Any feedbacks and ideas are welcome. Thank you in advance!
> >
> > Hyeonggon
> >
> > https://lore.kernel.org/linux-mm/202307172140.3b34825a-oliver.sang@xxxxxxxxx
> > [1]
> > https://lore.kernel.org/linux-mm/CAB=+i9S6Ykp90+4N1kCE=hiTJTE4wzJDi8k5pBjjO_3sf0aeqg@xxxxxxxxxxxxxx
> > [2]
> >
> > Hyeonggon Yoo (2):
> >   Revert "mm, slub: change percpu partial accounting from objects to
> >     pages"
> >   mm/slub: prefer NUMA locality over slight memory saving on NUMA
> >     machines
> >
> >  include/linux/slub_def.h |  2 --
> >  mm/slab.h                |  6 ++++
> >  mm/slub.c                | 76 ++++++++++++++++++++++++++----------
> > ----
> >  3 files changed, 55 insertions(+), 29 deletions(-)
> >
>