lkml.org 
[lkml]   [2006]   [Jul]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRE: [PATCH] mm: moving dirty pages balancing to pdfludh entirely
Date
From
Antonio Vargas writes:
> Maybe we should keep the sync-write logic if there is only one online
> cpu, and thus avoiding extra context switches when they are not
> profitable?

A parallelism makes sense even if 1 cpu and 1 user task is there because
of IO.
From other hand if user thread actually does inodes write back, it may
wait a lot (fs, jorn, io queue) events and get context switch.
The results of 3-4 repeated runs of "/usr/bin/time -f "%c"
iozone -i 0 -r 4 -s 1200m " in Pentium-4HT with 1GB RAM show that the
patch is useful for 1 cpu as well as for 2:

Old_1cpu new_1cpu
old_2cpu new_2cpu
/usr/bin/time %c 1932-3400 1700-3003
1900-2728 2014-2700
'vmstat 1' (cs/sec) 506-621 693-753 708-715
679-752
throughput(MB/sec) 54-58 71-94 53-59
74-105

Leonid

-----Original Message-----
From: Antonio Vargas [mailto:windenntw@gmail.com]
Sent: Monday, July 03, 2006 9:32 PM
To: Nikita Danilov; Ananiev, Leonid I; Linux Kernel Mailing List
Subject: Re: [PATCH] mm: moving dirty pages balancing to pdfludh
entirely
On 6/28/06, Nikita Danilov <nikita@clusterfs.com> wrote:
> Ananiev, Leonid I writes:
> > >From Leonid Ananiev
>
> Hello,
>
> >
> > Moving dirty pages balancing from user to kernel thread pdfludh
entirely
> > reduces extra long write(2) latencies, increases performance.
> >
>
> [...]
>
> > The benchmarks IOzone and Sysbench for file size 50% and 120%
of
> > RAM size on Pentium4, Xeon, Itanium have reported write and mix
> > throughput increasing about 25%. The described Iozone > 0.1 sec
write(2)
>
> Results are impressive.
>
> Wouldn't this interfere with current->backing_dev_info logic? This
field
> is set by __generic_file_aio_write_nolock() and checked by
> may_write_to_queue() to force heavy writes to do more pageout. Maybe
> pdflush threads should set this field too?
>
> > latencies are deleted. The condition writeback_in_progress() is
tested
> > earlier now. As a result extra pdflush works are not created and
number
> > of context switches increasing is inside variation limites.
>
> Nikita.

Maybe we should keep the sync-write logic if there is only one online
cpu, and thus avoiding extra context switches when they are not
profitable?

--
Greetz, Antonio Vargas aka winden of network
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2006-07-04 10:40    [from the cache]
©2003-2011 Jasper Spaans. Advertise on this site