[PATCH v2 0/4] Improve numa load balancing

[Srikar Dronamraju]

Here is a patchset that improves numa load balancing. It prefers
numa over cache during regular balancing. It also takes imbalance_pct into
consideration in numa_has_capacity() and provides few fixes to
task_numa_migrate().

Here are results from 5 runs of autonuma-benchmark
(https://github.com/pholasek/autonuma-benchmark)

KernelVersion: 4.1.0-rc7-tip
	Testcase:         Min         Max         Avg      StdDev
  elapsed_numa01:      858.85      949.18      915.64       33.06
  elapsed_numa02:       23.09       29.89       26.43        2.18
	Testcase:         Min         Max         Avg      StdDev
   system_numa01:     1516.72     1855.08     1686.24      113.95
   system_numa02:       63.69       79.06       70.35        5.87
	Testcase:         Min         Max         Avg      StdDev
     user_numa01:    73284.76    80818.21    78060.88     2773.60
     user_numa02:     1690.18     2071.07     1821.64      140.25
	Testcase:         Min         Max         Avg      StdDev
    total_numa01:    74801.50    82572.60    79747.12     2875.61
    total_numa02:     1753.87     2142.77     1891.99      143.59

KernelVersion: 4.1.0-rc7-tip + patchset

	Testcase:         Min         Max         Avg      StdDev     %Change
  elapsed_numa01:      718.47      916.32      867.63       75.47       5.24%
  elapsed_numa02:       19.12       28.79       24.78        3.46       5.73%
	Testcase:         Min         Max         Avg      StdDev     %Change
   system_numa01:      870.48     1493.04     1208.64      202.34      31.99%
   system_numa02:       41.25       64.56       55.12        7.96      23.59%
	Testcase:         Min         Max         Avg      StdDev     %Change
     user_numa01:    57736.89    74336.14    70092.39     6213.45      10.72%
     user_numa02:     1478.86     2122.50     1768.01      221.06       2.53%
	Testcase:         Min         Max         Avg      StdDev     %Change
    total_numa01:    58607.40    75628.90    71301.04     6380.18      11.17%
    total_numa02:     1520.11     2183.44     1823.13      227.16       3.15%




 Performance counter stats for 'system wide':

numa01
KernelVersion: 4.1.0-rc7-tip
          7,58,804      cs                                                           [100.00%]
          1,29,961      migrations                                                   [100.00%]
          7,08,643      faults
    3,97,97,92,613      cache-misses

     949.188997751 seconds time elapsed

KernelVersion: 4.1.0-rc7-tip + patchset
         11,38,993      cs                                                           [100.00%]
          2,05,490      migrations                                                   [100.00%]
          8,72,140      faults
    3,97,61,99,522      cache-misses

     907.583509759 seconds time elapsed


numa02
KernelVersion: 4.1.0-rc7-tip
            29,573      cs                                                           [100.00%]
             4,671      migrations                                                   [100.00%]
            33,236      faults
      12,68,88,021      cache-misses

      29.897733416 seconds time elapsed

KernelVersion: 4.1.0-rc7-tip + patchset
            23,422      cs                                                           [100.00%]
             4,306      migrations                                                   [100.00%]
            23,858      faults
       9,98,87,326      cache-misses

      22.776749102 seconds time elapsed

# numactl -H
available: 4 nodes (0-3)
node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
node 0 size: 32425 MB
node 0 free: 31352 MB
node 1 cpus: 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
node 1 size: 31711 MB
node 1 free: 30147 MB
node 2 cpus: 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71
node 2 size: 30431 MB
node 2 free: 29636 MB
node 3 cpus: 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
node 3 size: 32219 MB
node 3 free: 31053 MB
node distances:
node   0   1   2   3
  0:  10  20  40  40
  1:  20  10  40  40
  2:  40  40  10  20
  3:  40  40  20  10

Srikar Dronamraju (4):
  sched/tip:Prefer numa hotness over cache hotness
  sched: Consider imbalance_pct when comparing loads in numa_has_capacity
  sched: Fix task_numa_migrate to always update preferred node
  sched: Use correct nid while evaluating task weights

 kernel/sched/fair.c     | 109 ++++++++++++++----------------------------------
 kernel/sched/features.h |  18 +++-----
 2 files changed, 37 insertions(+), 90 deletions(-)

--
1.8.3.1

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