From: David Howells <dhowells@redhat.com>
To: Tejun Heo <tj@kernel.org>
Cc: dhowells@redhat.com, torvalds@linux-foundation.org,
mingo@elte.hu, peterz@infradead.org, awalls@radix.net,
linux-kernel@vger.kernel.org, jeff@garzik.org,
akpm@linux-foundation.org, jens.axboe@oracle.com,
rusty@rustcorp.com.au, cl@linux-foundation.org,
arjan@linux.intel.com, avi@redhat.com, johannes@sipsolutions.net,
andi@firstfloor.org, oleg@redhat.com
Subject: Re: [PATCHSET] workqueue: concurrency managed workqueue, take#4
Date: Tue, 16 Mar 2010 17:18:31 +0000 [thread overview]
Message-ID: <18911.1268759911@redhat.com> (raw)
In-Reply-To: <4B9FABB7.6030906@kernel.org>
Tejun Heo <tj@kernel.org> wrote:
> Sure, there could be a bug in the non-reentrance implementation but
> I'm leaning more towards a bug in work flushing before freeing thing
> which also seems to show up in the debugfs path. I'll try to
> reproduce the problem here and debug it.
I haven't managed to reproduce it since I reported it:-/
> That said, the numbers look generally favorable to CMWQ although the
> sample size is too small to draw conclusions. I'll try to get things
> fixed up so that testing can be smoother.
You have to take the numbers with a large pinch of salt, I think, in both
cases. Pulling over the otherwise unladen GigE network from the server with
the data in RAM is somewhat faster than sucking from disk. Furthermore, since
the test is massively parallel, with each thread reading separate data, the
result is going to be very much dependent on what order the reads happen to be
issued this time compared to the order they were issued when the cache was
filled.
I need to fix my slow-test server that's dangling at the end of an ethernet
over mains connection. That gives much more consistent results as the disk
speed is greater than the network connection speed.
Looking at the numbers, I think CMWQ may appear to give better results in the
cold-cache case by starting off confining many accesses to the cache to a
single CPU, given that cache object creation and data storage is done
asynchronously in the background. This is due to object creation getting
deferred until index creation is achieved (several loopups, mkdirs and
setxattrs), and then all dumped at once onto the CPU that handled the index
creation, as we discussed elsewhere.
The program I'm using to read the data doesn't give any real penalty when its
threads can't actually run in parallel, so it probably doesn't mind being
largely confined to the other CPU. But that's benchmarking for you...
You should probably also disregard the coldish-server numbers. I'm not sure
my desktop machine (which was acting as the server) was purged of the dataset.
I'd need to reboot the server to be sure, but that's inconvenient of my
desktop.
But, at a glance, the numbers don't appear to be too different. There are
cases where CMWQ definitely appears better, and some where it definitely
appears worse, but the spread is so huge, that could just be noise.
David
next prev parent reply other threads:[~2010-03-16 17:20 UTC|newest]
Thread overview: 86+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-02-26 12:22 [PATCHSET] workqueue: concurrency managed workqueue, take#4 Tejun Heo
2010-02-26 12:22 ` [PATCH 01/43] sched: consult online mask instead of active in select_fallback_rq() Tejun Heo
2010-02-26 12:22 ` [PATCH 02/43] sched: rename preempt_notifiers to sched_notifiers and refactor implementation Tejun Heo
2010-02-26 12:22 ` [PATCH 03/43] sched: refactor try_to_wake_up() Tejun Heo
2010-02-26 12:22 ` [PATCH 04/43] sched: implement __set_cpus_allowed() Tejun Heo
2010-02-26 12:22 ` [PATCH 05/43] sched: make sched_notifiers unconditional Tejun Heo
2010-02-26 12:22 ` [PATCH 06/43] sched: add wakeup/sleep sched_notifiers and allow NULL notifier ops Tejun Heo
2010-02-26 12:22 ` [PATCH 07/43] sched: implement try_to_wake_up_local() Tejun Heo
2010-02-28 12:33 ` Oleg Nesterov
2010-03-01 14:22 ` Tejun Heo
2010-02-26 12:22 ` [PATCH 08/43] workqueue: change cancel_work_sync() to clear work->data Tejun Heo
2010-02-26 12:22 ` [PATCH 09/43] acpi: use queue_work_on() instead of binding workqueue worker to cpu0 Tejun Heo
2010-02-26 12:22 ` [PATCH 10/43] stop_machine: reimplement without using workqueue Tejun Heo
2010-02-28 14:11 ` Oleg Nesterov
2010-03-01 15:07 ` Tejun Heo
2010-03-01 15:37 ` Oleg Nesterov
2010-03-01 16:36 ` Tejun Heo
2010-03-01 16:50 ` Oleg Nesterov
2010-03-01 18:02 ` Tejun Heo
2010-02-28 14:34 ` Oleg Nesterov
2010-03-01 15:11 ` Tejun Heo
2010-03-01 15:41 ` Oleg Nesterov
2010-02-26 12:22 ` [PATCH 11/43] workqueue: misc/cosmetic updates Tejun Heo
2010-02-26 12:22 ` [PATCH 12/43] workqueue: merge feature parameters into flags Tejun Heo
2010-02-26 12:22 ` [PATCH 13/43] workqueue: define masks for work flags and conditionalize STATIC flags Tejun Heo
2010-02-26 12:22 ` [PATCH 14/43] workqueue: separate out process_one_work() Tejun Heo
2010-02-26 12:22 ` [PATCH 15/43] workqueue: temporarily disable workqueue tracing Tejun Heo
2010-02-26 12:22 ` [PATCH 16/43] workqueue: kill cpu_populated_map Tejun Heo
2010-02-28 16:00 ` Oleg Nesterov
2010-03-01 15:32 ` Tejun Heo
2010-03-01 15:50 ` Oleg Nesterov
2010-03-01 16:19 ` Tejun Heo
2010-02-26 12:22 ` [PATCH 17/43] workqueue: update cwq alignement Tejun Heo
2010-02-28 17:12 ` Oleg Nesterov
2010-03-01 16:40 ` Tejun Heo
2010-02-26 12:22 ` [PATCH 18/43] workqueue: reimplement workqueue flushing using color coded works Tejun Heo
2010-02-28 20:31 ` Oleg Nesterov
2010-03-01 17:33 ` Tejun Heo
2010-03-01 19:47 ` Oleg Nesterov
2010-02-26 12:22 ` [PATCH 19/43] workqueue: introduce worker Tejun Heo
2010-02-26 12:22 ` [PATCH 20/43] workqueue: reimplement work flushing using linked works Tejun Heo
2010-03-01 14:53 ` Oleg Nesterov
2010-03-01 18:00 ` Tejun Heo
2010-03-01 18:51 ` Oleg Nesterov
2010-02-26 12:22 ` [PATCH 21/43] workqueue: implement per-cwq active work limit Tejun Heo
2010-02-26 12:22 ` [PATCH 22/43] workqueue: reimplement workqueue freeze using max_active Tejun Heo
2010-02-26 12:23 ` [PATCH 23/43] workqueue: introduce global cwq and unify cwq locks Tejun Heo
2010-02-26 12:23 ` [PATCH 24/43] workqueue: implement worker states Tejun Heo
2010-02-26 12:23 ` [PATCH 25/43] workqueue: reimplement CPU hotplugging support using trustee Tejun Heo
2010-02-26 12:23 ` [PATCH 26/43] workqueue: make single thread workqueue shared worker pool friendly Tejun Heo
2010-02-26 12:23 ` [PATCH 27/43] workqueue: add find_worker_executing_work() and track current_cwq Tejun Heo
2010-02-26 12:23 ` [PATCH 28/43] workqueue: carry cpu number in work data once execution starts Tejun Heo
2010-02-28 7:08 ` [PATCH UPDATED " Tejun Heo
2010-02-26 12:23 ` [PATCH 29/43] workqueue: implement WQ_NON_REENTRANT Tejun Heo
2010-02-26 12:23 ` [PATCH 30/43] workqueue: use shared worklist and pool all workers per cpu Tejun Heo
2010-02-26 12:23 ` [PATCH 31/43] workqueue: implement concurrency managed dynamic worker pool Tejun Heo
2010-02-26 12:23 ` [PATCH 32/43] workqueue: increase max_active of keventd and kill current_is_keventd() Tejun Heo
2010-02-26 12:23 ` [PATCH 33/43] workqueue: add system_wq, system_long_wq and system_nrt_wq Tejun Heo
2010-02-26 12:23 ` [PATCH 34/43] workqueue: implement DEBUGFS/workqueue Tejun Heo
2010-02-28 7:13 ` [PATCH UPDATED " Tejun Heo
2010-02-26 12:23 ` [PATCH 35/43] workqueue: implement several utility APIs Tejun Heo
2010-02-28 7:15 ` [PATCH UPDATED " Tejun Heo
2010-02-26 12:23 ` [PATCH 36/43] libata: take advantage of cmwq and remove concurrency limitations Tejun Heo
2010-02-26 12:23 ` [PATCH 37/43] async: use workqueue for worker pool Tejun Heo
2010-02-26 12:23 ` [PATCH 38/43] fscache: convert object to use workqueue instead of slow-work Tejun Heo
2010-02-26 12:23 ` [PATCH 39/43] fscache: convert operation " Tejun Heo
2010-02-26 12:23 ` [PATCH 40/43] fscache: drop references to slow-work Tejun Heo
2010-02-26 12:23 ` [PATCH 41/43] cifs: use workqueue instead of slow-work Tejun Heo
2010-02-28 7:09 ` [PATCH UPDATED " Tejun Heo
2010-02-26 12:23 ` [PATCH 42/43] gfs2: " Tejun Heo
2010-02-28 7:10 ` [PATCH UPDATED " Tejun Heo
2010-02-26 12:23 ` [PATCH 43/43] slow-work: kill it Tejun Heo
2010-02-27 22:52 ` [PATCH] workqueue: Fix build on PowerPC Anton Blanchard
2010-02-28 6:08 ` Tejun Heo
2010-02-28 1:00 ` [PATCH] workqueue: Fix a compile warning in work_busy Anton Blanchard
2010-02-28 6:18 ` Tejun Heo
2010-02-28 1:11 ` [PATCHSET] workqueue: concurrency managed workqueue, take#4 Anton Blanchard
2010-02-28 6:32 ` Tejun Heo
2010-03-10 14:52 ` David Howells
2010-03-12 5:03 ` Tejun Heo
2010-03-12 11:23 ` David Howells
2010-03-12 22:55 ` Tejun Heo
2010-03-16 14:38 ` David Howells
2010-03-16 16:03 ` Tejun Heo
2010-03-16 17:18 ` David Howells [this message]
2010-04-25 8:09 ` [PATCHSET UPDATED] " Tejun Heo
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=18911.1268759911@redhat.com \
--to=dhowells@redhat.com \
--cc=akpm@linux-foundation.org \
--cc=andi@firstfloor.org \
--cc=arjan@linux.intel.com \
--cc=avi@redhat.com \
--cc=awalls@radix.net \
--cc=cl@linux-foundation.org \
--cc=jeff@garzik.org \
--cc=jens.axboe@oracle.com \
--cc=johannes@sipsolutions.net \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=oleg@redhat.com \
--cc=peterz@infradead.org \
--cc=rusty@rustcorp.com.au \
--cc=tj@kernel.org \
--cc=torvalds@linux-foundation.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