Re: [PATCH] sched/fair: Rate limit calls to update_blocked_averages() for NOHZ

[Vincent Guittot]

On Wed, 7 Apr 2021 at 19:19, Tim Chen <tim.c.chen@xxxxxxxxxxxxxxx> wrote:
>
>
>
> On 4/7/21 7:02 AM, Vincent Guittot wrote:
> > Hi Tim,
> >
> > On Wed, 24 Mar 2021 at 17:05, Tim Chen <tim.c.chen@xxxxxxxxxxxxxxx> wrote:
> >>
> >>
> >>
> >> On 3/24/21 6:44 AM, Vincent Guittot wrote:
> >>> Hi Tim,
> >>
> >>>
> >>> IIUC your problem, we call update_blocked_averages() but because of:
> >>>
> >>>               if (this_rq->avg_idle < curr_cost + sd->max_newidle_lb_cost) {
> >>>                       update_next_balance(sd, &next_balance);
> >>>                       break;
> >>>               }
> >>>
> >>> the for_each_domain loop stops even before running load_balance on the 1st
> >>> sched domain level which means that update_blocked_averages() was called
> >>> unnecessarily.
> >>>
> >>
> >> That's right
> >>
> >>> And this is even more true with a small sysctl_sched_migration_cost which allows newly
> >>> idle LB for very small this_rq->avg_idle. We could wonder why you set such a low value
> >>> for sysctl_sched_migration_cost which is lower than the max_newidle_lb_cost of the
> >>> smallest domain but that's probably because of task_hot().
> >>>
> >>> if avg_idle is lower than the sd->max_newidle_lb_cost of the 1st sched_domain, we should
> >>> skip spin_unlock/lock and for_each_domain() loop entirely
> >>>
> >>> Maybe something like below:
> >>>
> >>
> >> The patch makes sense.  I'll ask our benchmark team to queue this patch for testing.
> >
> > Do you have feedback from your benchmark team ?
> >
>
> Vincent,
>
> Thanks for following up. I just got some data back from the benchmark team.
> The performance didn't change with your patch.  And the overall cpu% of update_blocked_averages
> also remain at about the same level.  My first thought was perhaps this update
> still didn't catch all the calls to update_blocked_averages
>
>         if (this_rq->avg_idle < sysctl_sched_migration_cost ||
> -           !READ_ONCE(this_rq->rd->overload)) {
> +           !READ_ONCE(this_rq->rd->overload) ||
> +           (sd && this_rq->avg_idle < sd->max_newidle_lb_cost)) {
>
> To experiment, I added one more check on the next_balance to further limit
> the path to actually do idle load balance with the next_balance time.
>
>         if (this_rq->avg_idle < sysctl_sched_migration_cost ||
> -           !READ_ONCE(this_rq->rd->overload)) {
> +           time_before(jiffies, this_rq->next_balance) ||
> +           !READ_ONCE(this_rq->rd->overload) ||
> +           (sd && this_rq->avg_idle < sd->max_newidle_lb_cost)) {
>
> 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 ?

> 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 ?

Vincent

> have a better idea.  I won't have access to the test system and workload
> till probably next week.
>
> Thanks.
>
> Tim