[<prev] [next>] [day] [month] [year] [list]
Message-ID: <20211208145223.64958-1-wanghonglei@didichuxing.com>
Date: Wed, 8 Dec 2021 22:52:23 +0800
From: Honglei Wang <wanghonglei@...ichuxing.com>
To: Ingo Molnar <mingo@...hat.com>,
Peter Zijlstra <peterz@...radead.org>,
Juri Lelli <juri.lelli@...hat.com>,
Vincent Guittot <vincent.guittot@...aro.org>,
Dietmar Eggemann <dietmar.eggemann@....com>,
Steven Rostedt <rostedt@...dmis.org>,
Ben Segall <bsegall@...gle.com>,
"Mel Gorman" <mgorman@...e.de>,
Daniel Bristot de Oliveira <bristot@...hat.com>,
<linux-kernel@...r.kernel.org>, Jonathan Corbet <corbet@....net>
CC: Huaixin Chang <changhuaixin@...ux.alibaba.com>,
Honglei Wang <jameshongleiwang@....com>
Subject: [PATCH v2 3/3] sched/fair: update burst feature description
Update the burst feature description due to the burst periods
limit.
Signed-off-by: Honglei Wang <wanghonglei@...ichuxing.com>
---
Documentation/scheduler/sched-bwc.rst | 12 ++++++++++++
1 file changed, 12 insertions(+)
diff --git a/Documentation/scheduler/sched-bwc.rst b/Documentation/scheduler/sched-bwc.rst
index 173c14110c85..eac39d42ebf9 100644
--- a/Documentation/scheduler/sched-bwc.rst
+++ b/Documentation/scheduler/sched-bwc.rst
@@ -65,6 +65,18 @@ there many cgroups or CPU is under utilized, the interference is
limited. More details are shown in:
https://lore.kernel.org/lkml/5371BD36-55AE-4F71-B9D7-B86DC32E3D2B@linux.alibaba.com/
+We need to limit the burst periods to 2, because we found something interesting
+that with the feature, tasks have chance to get more cpu than quota in working
+periods. As every task could be scheduled out or sleep, the task group can get
+some more cpu for its burst workload in this way. But even if there's no burst
+workload, it can use those power to do daily jobs. If the average workload of a
+taskgroup is just a little bit higher than its quota, it can escape from the
+throttle and using the burst power for lots of periods.
+
+Permitting 2 periods for the burst cpu can help aovid such kind of 'stealing'
+and the task group won't lose its burst power if the periods refresh lands in
+the middle of a burst workloads.
+
Management
----------
Quota, period and burst are managed within the cpu subsystem via cgroupfs.
--
2.14.1
Powered by blists - more mailing lists