From: Tim Chen <tim.c.chen@linux.intel.com>
To: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Joel Fernandes <joel@joelfernandes.org>,
linux-kernel <linux-kernel@vger.kernel.org>,
Paul McKenney <paulmck@kernel.org>,
Frederic Weisbecker <fweisbec@gmail.com>,
Dietmar Eggeman <dietmar.eggemann@arm.com>,
Qais Yousef <qais.yousef@arm.com>,
Ben Segall <bsegall@google.com>,
Daniel Bristot de Oliveira <bristot@redhat.com>,
Ingo Molnar <mingo@redhat.com>,
Juri Lelli <juri.lelli@redhat.com>, Mel Gorman <mgorman@suse.de>,
Peter Zijlstra <peterz@infradead.org>,
Steven Rostedt <rostedt@goodmis.org>,
"Uladzislau Rezki (Sony)" <urezki@gmail.com>,
Neeraj upadhyay <neeraj.iitr10@gmail.com>,
Aubrey Li <aubrey.li@linux.intel.com>
Subject: Re: [PATCH] sched/fair: Rate limit calls to update_blocked_averages() for NOHZ
Date: Thu, 8 Apr 2021 16:05:00 -0700 [thread overview]
Message-ID: <4aa674d9-db49-83d5-356f-a20f9e2a7935@linux.intel.com> (raw)
In-Reply-To: <CAKfTPtCKavGWja42NdTmb+95ppG-WxYzoTJMmtgkCQcA-btfBw@mail.gmail.com>
On 4/8/21 7:51 AM, Vincent Guittot wrote:
>> I was suprised to find the overall cpu% consumption of update_blocked_averages
>> and throughput of the benchmark still didn't change much. So I took a
>> peek into the profile and found the update_blocked_averages calls shifted to the idle load balancer.
>> The call to update_locked_averages was reduced in newidle_balance so the patch did
>> what we intended. But the overall rate of calls to
>
> At least , we have removed the useless call to update_blocked_averages
> in newidle_balance when we will not perform any newly idle load
> balance
>
>> update_blocked_averages remain roughly the same, shifting from
>> newidle_balance to run_rebalance_domains.
>>
>> 100.00% (ffffffff810cf070)
>> |
>> ---update_blocked_averages
>> |
>> |--95.47%--run_rebalance_domains
>> | __do_softirq
>> | |
>> | |--94.27%--asm_call_irq_on_stack
>> | | do_softirq_own_stack
>
> The call of update_blocked_averages mainly comes from SCHED_SOFTIRQ.
> And as a result, not from the new path
> do_idle()->nohz_run_idle_balance() which has been added by this patch
> to defer the call to update_nohz_stats() after newlyidle_balance and
> before entering idle.
>
>> | | |
>> | | |--93.74%--irq_exit_rcu
>> | | | |
>> | | | |--88.20%--sysvec_apic_timer_interrupt
>> | | | | asm_sysvec_apic_timer_interrupt
>> | | | | |
>> ...
>> |
>> |
>> --4.53%--newidle_balance
>> pick_next_task_fair
>>
>> I was expecting idle load balancer to be rate limited to 60 Hz, which
>
> Why 60Hz ?
>
My thinking is we will trigger load balance only after rq->next_balance.
void trigger_load_balance(struct rq *rq)
{
/* Don't need to rebalance while attached to NULL domain */
if (unlikely(on_null_domain(rq)))
return;
if (time_after_eq(jiffies, rq->next_balance))
raise_softirq(SCHED_SOFTIRQ);
nohz_balancer_kick(rq);
}
And it seems like next_balance is set to be 60 Hz
static void rebalance_domains(struct rq *rq, enum cpu_idle_type idle)
{
int continue_balancing = 1;
int cpu = rq->cpu;
int busy = idle != CPU_IDLE && !sched_idle_cpu(cpu);
unsigned long interval;
struct sched_domain *sd;
/* Earliest time when we have to do rebalance again */
unsigned long next_balance = jiffies + 60*HZ;
>> should be 15 jiffies apart on the test system with CONFIG_HZ_250.
>> When I did a trace on a single CPU, I see that update_blocked_averages
>> are often called between 1 to 4 jiffies apart, which is at a much higher
>> rate than I expected. I haven't taken a closer look yet. But you may
>
> 2 things can trigger a SCHED_SOFTIRQ/run_rebalance_domains:
> - the need for an update of blocked load which should not happen more
> than once every 32ms which means a rate of around 30Hz
> - the need for a load balance of a sched_domain. The min interval for
> a sched_domain is its weight when the CPU is idle which is usually few
> jiffies
>
> The only idea that I have for now is that we spend less time in
> newidle_balance which changes the dynamic of your system.
>
> In your trace, could you check if update_blocked_averages is called
> during the tick ? and Is the current task idle task ?
Here's a snapshot of the trace. However I didn't have the current task in my trace.
You can tell the frequency that update_blocked_averages is called on
cpu 2 by the jiffies value. They are quite close together (1 to 3 jiffies apart).
When I have a chance to get on the machine, I'll take another look
at the current task and whether we got to trigger_load_balance() from scheduler_tick().
3.505 ( ): probe:update_blocked_averages:(ffffffff810cf070) cpu=2 jiffies=0x1004fb731
4.505 ( ): probe:update_blocked_averages:(ffffffff810cf070) cpu=2 jiffies=0x1004fb732
6.484 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb733
6.506 ( ): probe:update_blocked_averages:(ffffffff810cf070) cpu=2 jiffies=0x1004fb734
9.503 ( ): probe:update_blocked_averages:(ffffffff810cf070) cpu=2 jiffies=0x1004fb737
11.504 ( ): probe:update_blocked_averages:(ffffffff810cf070) cpu=2 jiffies=0x1004fb739
11.602 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb76c jiffies=0x1004fb739
11.624 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb739
11.642 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb739
11.645 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb739
11.977 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb739
12.003 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb739
12.015 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb739
12.043 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb739
12.567 ( ): probe:update_blocked_averages:(ffffffff810cf070) cpu=2 jiffies=0x1004fb73a
13.856 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb76c jiffies=0x1004fb73b
13.910 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73b
14.003 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73b
14.159 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73b
14.203 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73b
14.223 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73b
14.301 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73b
14.504 ( ): probe:update_blocked_averages:(ffffffff810cf070) cpu=2 jiffies=0x1004fb73c
14.637 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb76c jiffies=0x1004fb73c
14.666 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73c
15.059 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73c
15.083 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73c
15.100 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73c
15.103 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73c
15.150 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73c
15.227 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73c
15.248 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73c
15.311 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73c
15.503 ( ): probe:update_blocked_averages:(ffffffff810cf070) cpu=2 jiffies=0x1004fb73d
16.140 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb76c jiffies=0x1004fb73d
16.185 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73d
16.224 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73d
16.340 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73d
16.384 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73d
16.503 ( ): probe:update_blocked_averages:(ffffffff810cf070) cpu=2 jiffies=0x1004fb73e
16.993 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb76c jiffies=0x1004fb73e
17.504 ( ): probe:update_blocked_averages:(ffffffff810cf070) cpu=2 jiffies=0x1004fb73f
17.630 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb76c jiffies=0x1004fb73f
17.830 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73f
18.015 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73f
18.031 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73f
18.036 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73f
18.040 ( ): probe:newidle_balance:(ffffffff810d2470) this_rq=0xffff88fe7f8aae00 next_balance=0x1004fb731 jiffies=0x1004fb73f
18.502 ( ): probe:update_blocked_averages:(ffffffff810cf070) cpu=2 jiffies=0x1004fb740
Thanks for taking a look.
Tim
next prev parent reply other threads:[~2021-04-08 23:05 UTC|newest]
Thread overview: 44+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-01-22 15:46 [PATCH] sched/fair: Rate limit calls to update_blocked_averages() for NOHZ Joel Fernandes (Google)
2021-01-22 16:56 ` Vincent Guittot
2021-01-22 18:39 ` Qais Yousef
2021-01-22 19:14 ` Joel Fernandes
2021-01-25 13:23 ` Vincent Guittot
2021-01-26 16:36 ` Qais Yousef
2021-01-22 19:10 ` Joel Fernandes
2021-01-25 10:44 ` Dietmar Eggemann
2021-01-25 17:30 ` Vincent Guittot
2021-01-25 17:53 ` Dietmar Eggemann
2021-01-25 14:42 ` Vincent Guittot
2021-01-27 18:43 ` Joel Fernandes
2021-01-28 13:57 ` Vincent Guittot
2021-01-28 15:09 ` Joel Fernandes
2021-01-28 16:57 ` Qais Yousef
[not found] ` <CAKfTPtBvwm9vZb5C=2oTF6N-Ht6Rvip4Lv18yi7O3G8e-_ZWdg@mail.gmail.com>
2021-01-29 17:27 ` Vincent Guittot
2021-02-03 11:54 ` Dietmar Eggemann
2021-02-03 13:12 ` Vincent Guittot
2021-02-04 9:47 ` Dietmar Eggemann
2021-02-03 17:09 ` Qais Yousef
2021-02-03 17:35 ` Vincent Guittot
2021-02-04 10:45 ` Qais Yousef
2021-02-03 19:56 ` Joel Fernandes
2021-03-23 21:37 ` Tim Chen
2021-03-24 13:44 ` Vincent Guittot
2021-03-24 16:05 ` Tim Chen
2021-04-07 14:02 ` Vincent Guittot
2021-04-07 17:19 ` Tim Chen
2021-04-08 14:51 ` Vincent Guittot
2021-04-08 23:05 ` Tim Chen [this message]
2021-04-09 15:26 ` Vincent Guittot
2021-04-09 17:59 ` Tim Chen
2021-05-10 21:59 ` Tim Chen
2021-05-11 15:25 ` Vincent Guittot
2021-05-11 17:25 ` Tim Chen
2021-05-11 17:56 ` Vincent Guittot
2021-05-12 13:59 ` Qais Yousef
2021-05-13 18:45 ` Tim Chen
2021-05-17 16:14 ` Qais Yousef
2021-06-11 20:00 ` Tim Chen
2021-06-18 10:28 ` Vincent Guittot
2021-06-18 16:14 ` Tim Chen
2021-06-25 8:50 ` Vincent Guittot
2021-02-01 15:13 ` Joel Fernandes
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4aa674d9-db49-83d5-356f-a20f9e2a7935@linux.intel.com \
--to=tim.c.chen@linux.intel.com \
--cc=aubrey.li@linux.intel.com \
--cc=bristot@redhat.com \
--cc=bsegall@google.com \
--cc=dietmar.eggemann@arm.com \
--cc=fweisbec@gmail.com \
--cc=joel@joelfernandes.org \
--cc=juri.lelli@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mgorman@suse.de \
--cc=mingo@redhat.com \
--cc=neeraj.iitr10@gmail.com \
--cc=paulmck@kernel.org \
--cc=peterz@infradead.org \
--cc=qais.yousef@arm.com \
--cc=rostedt@goodmis.org \
--cc=urezki@gmail.com \
--cc=vincent.guittot@linaro.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).