]> git.karo-electronics.de Git - karo-tx-linux.git/commit
per-task block plug can reduce block queue lock contention and increase
authorShaohua Li <shaohua.li@intel.com>
Wed, 24 Aug 2011 23:46:48 +0000 (09:46 +1000)
committerStephen Rothwell <sfr@canb.auug.org.au>
Thu, 22 Sep 2011 08:26:20 +0000 (18:26 +1000)
commit8a94b99c119d9e6d3643407553b048c3c4a09ded
tree1f81ed581449accb8eccebb2b725f5b27e1cd7e2
parentfa6e792ec3281eaa4971f854257efba4b32da166
per-task block plug can reduce block queue lock contention and increase
request merge.  Currently page reclaim doesn't support it.  I originally
thought page reclaim doesn't need it, because kswapd thread count is
limited and file cache write is done at flusher mostly.

When I test a workload with heavy swap in a 4-node machine, each CPU is
doing direct page reclaim and swap.  This causes block queue lock
contention.  In my test, without below patch, the CPU utilization is about
2% ~ 7%.  With the patch, the CPU utilization is about 1% ~ 3%.  Disk
throughput isn't changed.  This should improve normal kswapd write and
file cache write too (increase request merge for example), but might not
be so obvious as I explain above.

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Minchan Kim <minchan.kim@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
mm/vmscan.c