Files
tegra-linux-noble/include/linux
Dan Williams 8b3e6cdc53 md: introduce get_priority_stripe() to improve raid456 write performance
Improve write performance by preventing the delayed_list from dumping all its
stripes onto the handle_list in one shot.  Delayed stripes are now further
delayed by being held on the 'hold_list'.  The 'hold_list' is bypassed when:

  * a STRIPE_IO_STARTED stripe is found at the head of 'handle_list'
  * 'handle_list' is empty and i/o is being done to satisfy full stripe-width
    write requests
  * 'bypass_count' is less than 'bypass_threshold'.  By default the threshold
    is 1, i.e. every other stripe handled is a preread stripe provided the
    top two conditions are false.

Benchmark data:
System: 2x Xeon 5150, 4x SATA, mem=1GB
Baseline: 2.6.24-rc7
Configuration: mdadm --create /dev/md0 /dev/sd[b-e] -n 4 -l 5 --assume-clean
Test1: dd if=/dev/zero of=/dev/md0 bs=1024k count=2048
  * patched:  +33% (stripe_cache_size = 256), +25% (stripe_cache_size = 512)

Test2: tiobench --size 2048 --numruns 5 --block 4096 --block 131072 (XFS)
  * patched: +13%
  * patched + preread_bypass_threshold = 0: +37%

Changes since v1:
* reduce bypass_threshold from (chunk_size / sectors_per_chunk) to (1) and
  make it configurable.  This defaults to fairness and modest performance
  gains out of the box.
Changes since v2:
* [neilb@suse.de]: kill STRIPE_PRIO_HI and preread_needed as they are not
  necessary, the important change was clearing STRIPE_DELAYED in
  add_stripe_bio and this has been moved out to make_request for the hang
  fix.
* [neilb@suse.de]: simplify get_priority_stripe
* [dan.j.williams@intel.com]: reset the bypass_count when ->hold_list is
  sampled empty (+11%)
* [dan.j.williams@intel.com]: decrement the bypass_count at the detection
  of stripes being naturally promoted off of hold_list +2%.  Note, resetting
  bypass_count instead of decrementing on these events yields +4% but that is
  probably too aggressive.
Changes since v3:
* cosmetic fixups

Tested-by: James W. Laferriere <babydr@baby-dragons.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 08:58:42 -07:00
..
2008-02-19 10:04:00 +01:00
2008-03-04 13:41:23 +01:00
2008-04-19 09:59:43 +10:00
2008-04-22 15:16:32 -05:00
2008-04-17 20:05:38 +02:00
2008-03-04 16:35:12 -08:00
2008-02-14 20:58:05 -08:00
2008-03-04 14:47:06 -08:00
2008-04-21 11:22:28 -05:00
2008-04-25 13:26:55 +01:00
2008-04-25 13:26:55 +01:00
2008-02-08 02:09:56 +00:00
2008-04-25 13:26:55 +01:00
2008-04-21 22:38:45 +00:00
2008-02-08 09:22:24 -08:00
2008-04-28 08:58:30 -07:00
2008-04-19 19:19:54 +02:00
2008-04-28 08:58:35 -07:00
2008-04-22 11:34:59 +02:00
2008-04-04 18:36:49 +02:00
2008-04-17 12:22:31 +02:00
2008-04-23 21:25:36 -04:00
2008-03-17 22:48:46 -07:00
2008-02-19 21:00:18 +01:00
2008-04-19 19:10:28 -07:00
2008-04-26 17:36:37 +02:00
2008-04-17 20:05:42 +02:00
2008-04-19 19:44:57 +02:00
2008-04-25 00:46:09 -04:00
2008-04-28 08:58:32 -07:00
2008-04-15 21:22:17 +10:00
2008-02-26 14:03:47 +09:00
2008-03-04 16:35:15 -08:00
2008-02-08 09:22:31 -08:00
2008-02-14 21:13:33 -08:00
2008-04-28 08:58:29 -07:00
2008-04-19 16:54:56 -04:00
2008-04-19 16:55:29 -04:00
2008-02-14 21:13:33 -08:00
2008-04-20 21:47:03 -07:00
2008-04-20 21:47:03 -07:00
2008-04-20 21:47:07 -07:00
2008-02-07 23:11:56 -08:00
2008-03-17 22:46:46 -07:00
2008-02-08 09:22:31 -08:00
2008-02-08 09:22:26 -08:00
2008-02-08 09:22:41 -08:00
2008-04-17 10:43:01 -04:00
2008-04-25 00:25:08 +02:00
2008-02-07 08:42:34 -08:00
2008-04-28 08:58:32 -07:00
2008-04-15 13:26:52 -04:00
2008-02-07 08:42:30 -08:00
2008-02-08 09:22:27 -08:00
2008-02-07 08:42:34 -08:00
2008-04-28 08:58:20 -07:00
2008-04-28 08:58:29 -07:00
2008-04-19 19:44:59 +02:00
2008-04-17 20:05:37 +02:00
2008-04-17 20:05:36 +02:00
2008-04-24 21:16:57 -07:00
2008-04-07 13:14:22 -07:00
2008-04-28 08:58:21 -07:00
2008-03-06 15:30:40 -05:00