From: Mike Galbraith <efault@gmx.de>
To: Helge Hafting <helgehaf@aitel.hist.no>
Cc: Felipe Alfaro Solana <felipe_alfaro@linuxmail.org>,
davidm@hpl.hp.com, LKML <linux-kernel@vger.kernel.org>
Subject: Re: O(1) scheduler starvation
Date: Wed, 18 Jun 2003 14:32:14 +0200 [thread overview]
Message-ID: <5.2.0.9.2.20030618141902.02773c18@pop.gmx.net> (raw)
In-Reply-To: <3EF05816.6050307@aitel.hist.no>
At 02:16 PM 6/18/2003 +0200, Helge Hafting wrote:
>Mike Galbraith wrote:
>>At 09:53 AM 6/18/2003 +0200, Felipe Alfaro Solana wrote:
>>
>>>Hi!
>>>
>>>I've been poking around and found the following link on O(1) scheduler
>>>starvation problems:
>>>
>>>http://www.hpl.hp.com/research/linux/kernel/o1-starve.php
>>>
>>>The web page contains a small test program which supposedly is able to
>>>make two processes starvate. However, I've been unable to reproduce what
>>>is described in the above link. In fact, the CPU usage ratio ranges
>>>between 60-40% or 70-30% in the worst cases.
>>
>>(you're talking about with my monotinic_clock() diff in your kernel right?)
>>If you examine the priorities vs cpu usage, therein lies a big fat bug.
>>I think the fundamental problem is that you can only execute in series,
>>but can sleep in parallel, which makes for more sleep time existing than
>>all execution time combined.
>
>Would dividing the sleep time by the number of sleepers fix this?
>Or is division a too heavy operation here?
That won't work. You'd have to plunk it all into a pot, and divvy it up by
%cpu usage or something. I solved it the simple way. I keep a smoothed
run_avg (%cpu * 100), and use that as a sleep_avg limit... ie if you're
run_avg is 9000 (you're eating 10% cpu), no sleep will push you into
insanity land. I still have the problem that tasks will seek their
appropriate priority level and _stick_ there, so I probably need to add
some decay. (which will no doubt open the next can-O-worms)
-Mike
next prev parent reply other threads:[~2003-06-18 12:14 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2003-06-18 7:53 Felipe Alfaro Solana
2003-06-18 12:04 ` Mike Galbraith
2003-06-18 12:16 ` Helge Hafting
2003-06-18 12:32 ` Mike Galbraith [this message]
2003-06-18 14:22 ` Felipe Alfaro Solana
2003-06-18 15:54 ` Mike Galbraith
2003-06-18 15:59 ` Con Kolivas
2003-06-18 16:29 ` Mike Galbraith
2003-06-18 21:30 ` Felipe Alfaro Solana
2003-06-18 22:32 ` Mike Galbraith
2003-06-18 16:52 ` William Lee Irwin III
2003-06-18 17:17 ` Mike Galbraith
2003-06-18 16:51 ` William Lee Irwin III
2003-06-18 20:44 Ricardo Galli
2003-06-18 20:48 ` Marc-Christian Petersen
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=5.2.0.9.2.20030618141902.02773c18@pop.gmx.net \
--to=efault@gmx.de \
--cc=davidm@hpl.hp.com \
--cc=felipe_alfaro@linuxmail.org \
--cc=helgehaf@aitel.hist.no \
--cc=linux-kernel@vger.kernel.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
all inboxes | Powered by JetHome®