Re: [PATCH v9 12/13] memcg: create support routines for page writeback

[Wu Fengguang <fengguang.wu@intel.com>]

> > +
> > +			__set_current_state(TASK_UNINTERRUPTIBLE);
> > +			io_schedule_timeout(pause);
> > +
> 
> How do you think about MAX_PAUSE/PASS_GOOD ?
> ==
>                 /*
>                  * max-pause area. If dirty exceeded but still within this
>                  * area, no need to sleep for more than 200ms: (a) 8 pages per
>                  * 200ms is typically more than enough to curb heavy dirtiers;
>                  * (b) the pause time limit makes the dirtiers more responsive.
>                  */
>                 if (nr_dirty < dirty_thresh +
>                                dirty_thresh / DIRTY_MAXPAUSE_AREA &&
>                     time_after(jiffies, start_time + MAX_PAUSE))
>                         break;
>                 /*
>                  * pass-good area. When some bdi gets blocked (eg. NFS server
>                  * not responding), or write bandwidth dropped dramatically due
>                  * to concurrent reads, or dirty threshold suddenly dropped and
>                  * the dirty pages cannot be brought down anytime soon (eg. on
>                  * slow USB stick), at least let go of the good bdi's.
>                  */
>                 if (nr_dirty < dirty_thresh +
>                                dirty_thresh / DIRTY_PASSGOOD_AREA &&
>                     bdi_dirty < bdi_thresh)
>                         break;
> ==

Sorry that piece of code actually has some problems in JBOD setup.
I'm going to submit a patch for fixing it:

Subject: squeeze max-pause area and drop pass-good area
Date: Tue Aug 16 13:37:14 CST 2011

Remove the pass-good area introduced in ffd1f609ab10 ("writeback:
introduce max-pause and pass-good dirty limits") and make the
max-pause area smaller and safe.

This fixes ~30% performance regression in the ext3 data=writeback
fio_mmap_randwrite_64k/fio_mmap_randrw_64k test cases, where there are
12 JBOD disks, on each disk runs 8 concurrent tasks doing reads+writes.

Using deadline scheduler also has a regression, but not that big as
CFQ, so this suggests we have some write starvation.

The test logs show that

- the disks are sometimes under utilized

- global dirty pages sometimes rush high to the pass-good area for
  several hundred seconds, while in the mean time some bdi dirty pages
  drop to very low value (bdi_dirty << bdi_thresh).
  Then suddenly the global dirty pages dropped under global dirty
  threshold and bdi_dirty rush very high (for example, 2 times higher
  than bdi_thresh). During which time balance_dirty_pages() is not
  called at all.

So the problems are

1) The random writes progress so slow that they break the assumption of
the max-pause logic that "8 pages per 200ms is typically more than
enough to curb heavy dirtiers".

2) The max-pause logic ignored task_bdi_thresh and thus opens the
   possibility for some bdi's to over dirty pages, leading to
   (bdi_dirty >> bdi_thresh) and then (bdi_thresh >> bdi_dirty) for others.

3) The higher max-pause/pass-good thresholds somehow leads to some bad
   swing of dirty pages.

The fix is to allow the task to slightly dirty over task_bdi_thresh, but
no way to exceed bdi_dirty and/or global dirty_thresh.

Tests show that it fixed the JBOD regression completely (both behavior
and performance), while still being able to cut down large pause times
in balance_dirty_pages() for single-disk cases.

Reported-by: Li Shaohua <shaohua.li@intel.com>
Tested-by: Li Shaohua <shaohua.li@intel.com>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
---
 include/linux/writeback.h |   11 -----------
 mm/page-writeback.c       |   15 ++-------------
 2 files changed, 2 insertions(+), 24 deletions(-)

--- linux.orig/mm/page-writeback.c	2011-08-18 09:52:59.000000000 +0800
+++ linux/mm/page-writeback.c	2011-08-18 10:28:57.000000000 +0800
@@ -786,21 +786,10 @@ static void balance_dirty_pages(struct a
 		 * 200ms is typically more than enough to curb heavy dirtiers;
 		 * (b) the pause time limit makes the dirtiers more responsive.
 		 */
-		if (nr_dirty < dirty_thresh +
-			       dirty_thresh / DIRTY_MAXPAUSE_AREA &&
+		if (nr_dirty < dirty_thresh &&
+		    bdi_dirty < (task_bdi_thresh + bdi_thresh) / 2 &&
 		    time_after(jiffies, start_time + MAX_PAUSE))
 			break;
-		/*
-		 * pass-good area. When some bdi gets blocked (eg. NFS server
-		 * not responding), or write bandwidth dropped dramatically due
-		 * to concurrent reads, or dirty threshold suddenly dropped and
-		 * the dirty pages cannot be brought down anytime soon (eg. on
-		 * slow USB stick), at least let go of the good bdi's.
-		 */
-		if (nr_dirty < dirty_thresh +
-			       dirty_thresh / DIRTY_PASSGOOD_AREA &&
-		    bdi_dirty < bdi_thresh)
-			break;
 
 		/*
 		 * Increase the delay for each loop, up to our previous
--- linux.orig/include/linux/writeback.h	2011-08-16 23:34:27.000000000 +0800
+++ linux/include/linux/writeback.h	2011-08-18 09:53:03.000000000 +0800
@@ -12,15 +12,6 @@
  *
  *	(thresh - thresh/DIRTY_FULL_SCOPE, thresh)
  *
- * The 1/16 region above the global dirty limit will be put to maximum pauses:
- *
- *	(limit, limit + limit/DIRTY_MAXPAUSE_AREA)
- *
- * The 1/16 region above the max-pause region, dirty exceeded bdi's will be put
- * to loops:
- *
- *	(limit + limit/DIRTY_MAXPAUSE_AREA, limit + limit/DIRTY_PASSGOOD_AREA)
- *
  * Further beyond, all dirtier tasks will enter a loop waiting (possibly long
  * time) for the dirty pages to drop, unless written enough pages.
  *
@@ -31,8 +22,6 @@
  */
 #define DIRTY_SCOPE		8
 #define DIRTY_FULL_SCOPE	(DIRTY_SCOPE / 2)
-#define DIRTY_MAXPAUSE_AREA		16
-#define DIRTY_PASSGOOD_AREA		8
 
 /*
  * 4MB minimal write chunk size

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>