Message-Id: <20230613052004.2836135-1-void@manifault.com>
Date:   Tue, 13 Jun 2023 00:20:01 -0500
From:   David Vernet <void@...ifault.com>
To:     linux-kernel@...r.kernel.org
Cc:     mingo@...hat.com, peterz@...radead.org, juri.lelli@...hat.com,
        vincent.guittot@...aro.org, rostedt@...dmis.org,
        dietmar.eggemann@....com, bsegall@...gle.com, mgorman@...e.de,
        bristot@...hat.com, vschneid@...hat.com, joshdon@...gle.com,
        roman.gushchin@...ux.dev, tj@...nel.org, kernel-team@...a.com
Subject: [RFC PATCH 0/3] sched: Implement shared wakequeue in CFS

Overview
========

The scheduler must constantly strike a balance between work
conservation, and avoiding costly migrations which harm performance due
to e.g. decreased cache locality. The matter is further complicated by
the topology of the system. Migrating a task between cores on the same
LLC may be more optimal than keeping a task local to the CPU, whereas
migrating a task between LLCs or NUMA nodes may tip the balance in the
other direction.

With that in mind, while CFS is by and large mostly a work conserving
scheduler, there are certain instances where the scheduler will choose
to keep a task local to a CPU, when it would have been more optimal to
migrate it to an idle core.

An example of such a workload is the HHVM / web workload at Meta. HHVM
is a VM that JITs Hack and PHP code in service of web requests. Like
other JIT / compilation workloads, it tends to be heavily CPU bound, and
exhibit generally poor cache locality. To try and address this, we set
several debugfs (/sys/kernel/debug/sched) knobs on our HHVM workloads:

- migration_cost_ns -> 0
- latency_ns -> 20000000
- min_granularity_ns -> 10000000
- wakeup_granularity_ns -> 12000000

These knobs are intended both to encourage the scheduler to be as work
conserving as possible (migration_cost_ns -> 0), and also to keep tasks
running for relatively long time slices so as to avoid the overhead of
context switching (the other knobs). Collectively, these knobs provide a
substantial performance win; resulting in roughly a 20% improvement in
throughput. Worth noting, however, is that this improvement is _not_ at
full machine saturation.

That said, even with these knobs, we noticed that CPUs were still going
idle even when the host was overcommitted. In response, we wrote the
"shared wakequeue" (swqueue) feature proposed in this patch set. The
idea behind swqueue is simple: it enables the scheduler to be
aggressively work conserving by placing a waking task into a per-LLC
FIFO queue that can be pulled from by another core in the LLC FIFO queue
which can then be pulled from before it goes idle.

With this simple change, we were able to achieve a 1 - 1.6% improvement
in throughput, as well as a small, consistent improvement in p95 and p99
latencies, in HHVM. These performance improvements were in addition to
the wins from the debugfs knobs mentioned above.

Design
======

The design of swqueue is quite simple. An swqueue is simply a struct
list_head, and a spinlock:

struct swqueue {
	struct list_head list;
	spinlock_t lock;
} ____cacheline_aligned;

We create a struct swqueue per LLC, ensuring they're in their own
cachelines to avoid false sharing between CPUs on different LLCs.

When a task first wakes up, it enqueues itself in the swqueue of its
current LLC at the end of enqueue_task_fair(). Enqueues only happen if
the task was not manually migrated to the current core by
select_task_rq(), and is not pinned to a specific CPU.

A core will pull a task from its LLC's swqueue before calling
newidle_balance().

Difference between SIS_NODE
===========================

In [0] Peter proposed a patch that addresses Tejun's observations that
when workqueues are targeted towards a specific LLC on his Zen2 machine
with small CCXs, that there would be significant idle time due to
select_idle_sibling() not considering anything outside of the current
LLC.

This patch (SIS_NODE) is essentially the complement to the proposal
here. SID_NODE causes waking tasks to look for idle cores in neighboring
LLCs on the same die, whereas swqueue causes cores about to go idle to
look for enqueued tasks. That said, in its current form, the two
features at are a different scope as SIS_NODE searches for idle cores
between LLCs, while swqueue enqueues tasks within a single LLC.

The patch was since removed in [1], but we elect to compare its
performance to swqueue given that as described above, it's conceptually
complementary.

[0]: https://lore.kernel.org/all/20230530113249.GA156198@hirez.programming.kicks-ass.net/
[1]: https://lore.kernel.org/all/20230605175636.GA4253@hirez.programming.kicks-ass.net/

I observed that while SIS_NODE works quite well for hosts with small
CCXs, it can result in degraded performance on machines either with a
large number of total cores in a CCD, or for which the cache miss
penalty of migrating between CCXs is high, even on the same die.

For example, on Zen 4c hosts (Bergamo), CCXs within a CCD are muxed
through a single link to the IO die, and thus have similar cache miss
latencies as cores in remote CCDs.

Such subtleties could be taken into account with SIS_NODE, but
regardless, both features are conceptually complementary sides of the
same coin. SIS_NODE searches for idle cores for waking threads, whereas
swqueue searches for available work before a core goes idle.

Results
=======

Note that the motivation for the shared wakequeue feature was originally
arrived at using experiments in the sched_ext framework that's currently
being proposed upstream. The ~1 - 1.6% improvement in HHVM throughput
is similarly visible using work-conserving sched_ext schedulers (even
very simple ones like global FIFO).

In both single and multi socket / CCX hosts, this can measurably improve
performance. In addition to the performance gains observed on our
internal web workloads, we also observed an improvement in common
workloads such as kernel compile when running shared wakequeue. Here are
the results of running make -j$(nproc) built-in.a on several different
types of hosts configured with make allyesconfig on commit a27648c74210
("afs: Fix setting of mtime when creating a file/dir/symlink") on Linus'
tree (boost was disabled on all of these hosts when the experiments were
performed):

Single-socket | 32-core | 2-CCX | AMD 7950X Zen4

CPU max MHz:                     5879.8818
CPU min MHz:                     3000.0000
				o____________o_______o
				|    mean    | CPU   |
				o------------o-------o
NO_SWQUEUE + NO_SIS_NODE:	| 590.52s    | 3103% |
NO_SWQUEUE + SIS_NODE:		| 590.80s    | 3102% |
SWQUEUE + NO_SIS_NODE:		| 589.65s    | 3116% |
SWQUEUE + SIS_NODE:		| 589.99s    | 3115% |
				o------------o-------o

Takeaway: swqueue doesn't seem to provide a statistically significant
improvement for kernel compile on my 7950X. SIS_NODE similarly does not
have a noticeable effect on performance.

-------------------------------------------------------------------------------

Single-socket | 72-core | 6-CCX | AMD Milan Zen3

CPU max MHz:                     3245.0190
CPU min MHz:                     700.0000
				o_____________o_______o
				|    mean     | CPU   |
				o-------------o-------o
NO_SWQUEUE + NO_SIS_NODE:	| 1608.69s    | 6488% |
NO_SWQUEUE + SIS_NODE:		| 1610.24s    | 6473% |
SWQUEUE + NO_SIS_NODE:		| 1605.80s    | 6504% |
SWQUEUE + SIS_NODE:		| 1606.96s    | 6488% |
				o-------------o-------o

Takeaway: swqueue does provide a small statistically significant
improvement on Milan, but the compile times in general were quite long
relative to the 7950X Zen4, and the Bergamo Zen4c due to the lower clock
frequency. Milan also has larger CCXs than Bergamo, so it stands to
reason that select_idle_sibling() will have an easier time finding idle
cores inside the current CCX.

It also seems logical that SIS_NODE would hurt performance a bit here,
as all cores / CCXs are in the same NUMA node, so select_idle_sibling()
has to iterate over 72 cores; delaying task wakeup. That said, I'm not
sure that's a viable theory if total CPU% is lower with SIS_NODE.

-------------------------------------------------------------------------------

Single-socket | 176-core | 11-CCX | 2-CCX per CCD | AMD Bergamo Zen4c

CPU max MHz:                     1200.0000
CPU min MHz:                     1000.0000

				o____________o________o
				|    mean    | CPU    |
				o------------o--------o
NO_SWQUEUE + NO_SIS_NODE:	| 322.44s    | 15534% |
NO_SWQUEUE + SIS_NODE:		| 324.39s    | 15508% |
SWQUEUE + NO_SIS_NODE:		| 321.54s    | 15603% |
SWQUEUE + SIS_NODE:		| 321.88s    | 15622% |
				o------------o--------o

Takeaway: swqueue barely beats NO_SWQUEUE + NO_SIS_NODE, to the point
that it's arguably not statistically significant.
SIS_NODE results in a ~.9% performance degradation, for likely the same
reason as Milan: the host has a large number of LLCs within a single
socket, so task wakeup latencies suffer due to select_idle_node()
searching up to 11 CCXs.

Conclusion
==========

swqueue in this form seems to provide a small, but noticeable win for
front-end CPU-bound workloads spread over multiple CCXs. The reason
seems fairly straightforward: swqueue encourages work conservation
inside of a CCX by having a CPU do an O(1) pull from a per-LLC queue of
runnable tasks. As mentioned above, it is complementary to SIS_NODE,
which searches for idle cores on the wakeup path.

While swqueue in this form encourages work conservation, it of course
does not guarantee it given that we don't implement any kind of work
stealing between swqueues. In the future, we could potentially push CPU
utilization even higher by enabling work stealing between swqueues,
likely between CCXs on the same NUMA node.

Originally-by: Roman Gushchin <roman.gushchin@...ux.dev>
Signed-off-by: David Vernet <void@...ifault.com>

David Vernet (3):
  sched: Make migrate_task_to() take any task
  sched/fair: Add SWQUEUE sched feature and skeleton calls
  sched: Implement shared wakequeue in CFS

 include/linux/sched.h   |   2 +
 kernel/sched/core.c     |  52 ++++++-----
 kernel/sched/fair.c     | 200 +++++++++++++++++++++++++++++++++++++++-
 kernel/sched/features.h |   1 +
 kernel/sched/sched.h    |   6 +-
 5 files changed, 234 insertions(+), 27 deletions(-)

-- 
2.40.1

Powered by blists - more mailing lists