1 From: Jan Kara <jack@suse.cz>
2 Subject: Increase limits for starting writeback of dirty data
6 Return limits for dirty pages writeback to the numbers from SLES10. This dramatically
7 improves performance of workloads dirtying a lot of memory (e.g. simple databases not
8 using direct IO) and we're not aware it would harm anything.
10 Signed-off-by: Jan Kara <jack@suse.cz>
12 diff -rupX /home/jack/.kerndiffexclude linux-2.6.27/mm/page-writeback.c linux-2.6.27-dirty_limits//mm/page-writeback.c
13 --- linux-2.6.27/mm/page-writeback.c 2008-10-10 00:13:53.000000000 +0200
14 +++ linux-2.6.27-dirty_limits/mm/page-writeback.c 2009-01-07 16:09:24.000000000 +0100
15 @@ -66,7 +66,7 @@ static inline long sync_writeback_pages(
17 * Start background writeback (via pdflush) at this percentage
19 -int dirty_background_ratio = 5;
20 +int dirty_background_ratio = 10;
23 * free highmem will not be subtracted from the total free memory
24 @@ -77,7 +77,7 @@ int vm_highmem_is_dirtyable;
26 * The generator of dirty data starts writeback at this percentage
28 -int vm_dirty_ratio = 10;
29 +int vm_dirty_ratio = 40;
32 * The interval between `kupdate'-style writebacks, in jiffies