Re: [RFC 1/1] sched/fair: Consider asymmetric scheduler groups in load balancer

From: Tobias Huschle
Date: Mon Jun 05 2023 - 04:07:45 EST

On 2023-05-16 15:36, Vincent Guittot wrote:
On Mon, 15 May 2023 at 13:46, Tobias Huschle <huschle@xxxxxxxxxxxxx> wrote:

The current load balancer implementation implies that scheduler groups,
within the same domain, all host the same number of CPUs. This is
reflected in the condition, that a scheduler group, which is load
balancing and classified as having spare capacity, should pull work
from the busiest group, if the local group runs less processes than
the busiest one. This implies that these two groups should run the
same number of processes, which is problematic if the groups are not
of the same size.

The assumption that scheduler groups within the same scheduler domain
host the same number of CPUs appears to be true for non-s390
architectures. Nevertheless, s390 can have scheduler groups of unequal
size.

This introduces a performance degredation in the following scenario:

Consider a system with 8 CPUs, 6 CPUs are located on one CPU socket,
the remaining 2 are located on another socket:

Socket -----1----- -2-
CPU 1 2 3 4 5 6 7 8

Placing some workload ( x = one task ) yields the following
scenarios:

The first 5 tasks are distributed evenly across the two groups.

Socket -----1----- -2-
CPU 1 2 3 4 5 6 7 8
x x x x x

Adding a 6th task yields the following distribution:

Socket -----1----- -2-
CPU 1 2 3 4 5 6 7 8
SMT1 x x x x x
SMT2 x

Your description is a bit confusing for me. What you name CPU above
should be named Core, doesn' it ?

Could you share with us your scheduler topology ?


You are correct, it should say core instead of CPU.

One actual configuration from one of my test machines (lscpu -e):

CPU NODE DRAWER BOOK SOCKET CORE L1d:L1i:L2 ONLINE CONFIGURED POLARIZATION ADDRESS
0 0 0 0 0 0 0:0:0 yes yes horizontal 0
1 0 0 0 0 0 1:1:0 yes yes horizontal 1
2 0 0 0 0 1 2:2:0 yes yes horizontal 2
3 0 0 0 0 1 3:3:0 yes yes horizontal 3
4 0 0 0 0 2 4:4:0 yes yes horizontal 4
5 0 0 0 0 2 5:5:0 yes yes horizontal 5
6 0 0 0 0 3 6:6:0 yes yes horizontal 6
7 0 0 0 0 3 7:7:0 yes yes horizontal 7
8 0 0 0 0 4 8:8:0 yes yes horizontal 8
9 0 0 0 0 4 9:9:0 yes yes horizontal 9
10 0 0 0 0 5 10:10:0 yes yes horizontal 10
11 0 0 0 0 5 11:11:0 yes yes horizontal 11
12 0 0 0 1 6 12:12:0 yes yes horizontal 12
13 0 0 0 1 6 13:13:0 yes yes horizontal 13
14 0 0 0 1 7 14:14:0 yes yes horizontal 14
15 0 0 0 1 7 15:15:0 yes yes horizontal 15

So, 6 cores / 12 CPUs in one group 2 cores / 4 CPUs in the other.

If I run stress-ng with 8 cpu stressors on the original code I get a distribution
like this:

00 01 02 03 04 05 06 07 08 09 10 11 || 12 13 14 15
x x x x x x x x

Which means that the two cores in the smaller group are running into SMT while two
cores in the larger group are still idle. This is caused by the prefer_sibling path
which really wants to see both groups run the same number of tasks.


The task is added to the 2nd scheduler group, as the scheduler has the
assumption that scheduler groups are of the same size, so they should
also host the same number of tasks. This makes CPU 7 run into SMT
thread, which comes with a performance penalty. This means, that in
the window of 6-8 tasks, load balancing is done suboptimally, because
SMT is used although there is no reason to do so as fully idle CPUs
are still available.

Taking the weight of the scheduler groups into account, ensures that
a load balancing CPU within a smaller group will not try to pull tasks
from a bigger group while the bigger group still has idle CPUs
available.

Signed-off-by: Tobias Huschle <huschle@xxxxxxxxxxxxx>
---
kernel/sched/fair.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 48b6f0ca13ac..b1307d7e4065 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -10426,7 +10426,8 @@ static struct sched_group *find_busiest_group(struct lb_env *env)
* group's child domain.
*/
if (sds.prefer_sibling && local->group_type == group_has_spare &&
- busiest->sum_nr_running > local->sum_nr_running + 1)
+ busiest->sum_nr_running * local->group_weight >
+ local->sum_nr_running * busiest->group_weight + 1)

This is the prefer_sibling path. Could it be that you should disable
prefer_siling between your sockets for such topology ? the default
path compares the number of idle CPUs when groups has spare capacity



If I disable prefer_sibling (I played around with it a bit), I run into the problem,
that the tasks are distributed s.t. each group has the same amount of idle CPUs, which
yields distributions similar to this:

00 01 02 03 04 05 06 07 08 09 10 11 || 12 13 14 15
x x x x x x x x

Now both groups have 4 idle CPUs which fulfills the criteria imposed by the load balancer,
but the larger group is now running SMT while the smaller one is just idle.

So, in this asymmetric setup, both criteria seem to not work in an optimal way. Going for
the same number of idle CPUs or alternatively for the same number of running processes
both cause a sub-optimal distribution of tasks, leading to unnecessary SMT.

It seems also to be possible to address the regular load balancing path by aiming to have the
same unused capacity between groups instead of the same number of idle CPUs. This seems to
have been considered in the past, but the choice went in favor of the number of idle CPUs.
Since this decision was actively taken already, I focused on the prefer_sibling path.

The question now would be how to address this properly (or if I'm missing something here).
As mentioned in the cover letter, this was the most simplistic and least invasive approach
I could find, others might be more sophisticated but also have some side-effects.

I have a bit of a hard time leaving this one as-is, as it just introduces two additional
multiplications with no effect for most architectures. Maybe an architectures specific
inline function that the compiler can optimize away if not needed?

goto force_balance;

if (busiest->group_type != group_overloaded) {
--
2.34.1