Skip to contentSkip to conversation

b191f9b106e: -8.8% netperf.time.minor_page_faults

LKP Huang Ying 3 participants 3 messages Apr 1, 2015 – Apr 9, 2015

lore
Share & tools
lore

Huang Ying

author
#0 lore raw

FYI, we noticed the below changes on

git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
commit b191f9b106ea1a24a711dbebb2925d3313da5852 ("mm: numa: preserve PTE write permissions across a NUMA hinting fault")

testbox/testcase/testparams: lkp-ne02/netperf/performance-300s-200%-TCP_SENDFILE

bea66fbd11af1ca9 b191f9b106ea1a24a711dbebb2
---------------- --------------------------

         %stddev     %change         %stddev
             \          |                \  
     13115 ±  0%      -8.8%      11968 ±  0%  netperf.time.minor_page_faults
     65.63 ±  6%     -14.6%      56.05 ±  7%  netperf.time.user_time
     65.63 ±  6%     -14.6%      56.05 ±  7%  time.user_time
    728886 ± 12%     -29.2%     516121 ± 25%  proc-vmstat.pgalloc_dma32
       259 ± 12%     -21.8%        203 ± 19%  sched_debug.cfs_rq[10]:/.tg_load_contrib
       139 ± 26%     -40.9%         82 ± 48%  sched_debug.cfs_rq[10]:/.blocked_load_avg
   1437779 ±  5%      -9.7%    1298979 ±  2%  sched_debug.cfs_rq[11]:/.max_vruntime
   1437779 ±  5%      -9.7%    1298979 ±  2%  sched_debug.cfs_rq[11]:/.MIN_vruntime
       518 ± 37%     -52.1%        248 ± 40%  sched_debug.cfs_rq[3]:/.tg_load_contrib
       117 ±  6%     +13.5%        132 ±  9%  sched_debug.cfs_rq[3]:/.load
       191 ± 28%     +71.7%        328 ± 35%  sched_debug.cfs_rq[7]:/.tg_load_contrib
    306355 ± 47%     -33.0%     205153 ±  1%  sched_debug.cpu#11.nr_switches
    306423 ± 47%     -33.0%     205208 ±  1%  sched_debug.cpu#11.sched_count
      1050 ± 13%     +59.2%       1672 ± 18%  sched_debug.cpu#15.curr->pid
      1020 ±  6%     +35.1%       1379 ±  6%  sched_debug.cpu#4.curr->pid
      1082 ± 18%     +44.5%       1563 ± 28%  sched_debug.cpu#5.curr->pid
      1215 ± 14%     -17.6%       1001 ±  7%  sched_debug.cpu#8.curr->pid

lkp-ne02: Nehalem-EP
Memory: 5G

netperf.time.minor_page_faults

13400 ++---------------------------------*--------------------------------+

        |     *..      *..                :                                 |
  13200 *+. ..       ..   *..     .*..*.. :   *..*..*..*..*..     .*..*..   |
  13000 ++ *     *..*        *..*.       *                   *..*.       *..*
        |                                                                   |
  12800 ++                                                                  |
        |                                                                   |
  12600 ++                                                                  |
        |                                                                   |
  12400 ++                                                                  |
  12200 ++                                                                  |
        |                                                                   |
  12000 O+                         O          O     O  O     O              |
        |  O  O  O  O  O  O  O  O     O  O O     O        O     O           |

11800 ++------------------------------------------------------------------+

[*] bisect-good sample
[O] bisect-bad sample

To reproduce:

apt-get install ruby
git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git
cd lkp-tests
bin/setup-local job.yaml # the job file attached in this email
bin/run-local job.yaml

Disclaimer:
Results have been estimated based on internal Intel analysis and are provided for informational purposes only. Any difference in system hardware or software design or configuration may affect actual performance.

Thanks,
Ying Huang

_______________________________________________
LKP mailing list
LKP(a)linux.intel.com

---
testcase: netperf
default-monitors:
wait: pre-test
uptime:
iostat:
vmstat:
numa-numastat:
numa-vmstat:
numa-meminfo:
proc-vmstat:
proc-stat:
meminfo:
slabinfo:
interrupts:
lock_stat:
latency_stats:
softirqs:
bdi_dev_mapping:
diskstats:
nfsstat:
cpuidle:
cpufreq-stats:
turbostat:
pmeter:
sched_debug:
interval: 10
default-watchdogs:
watch-oom:
watchdog:
cpufreq_governor: performance
commit: e42391cd048809d903291d07f86ed3934ce138e9
model: Nehalem-EP
memory: 5G
nr_hdd_partitions: 12
hdd_partitions: "/dev/disk/by-id/scsi-35000c5000*part1"
swap_partitions: "/dev/disk/by-id/scsi-SATA_ST3500320NS_9QM6YWTY-part2"
runtime: 300s
nr_threads: 200%
perf-profile:
freq: 800
cluster: cs-localhost
if role server:
netserver:
if role client:
netperf:
test: TCP_SENDFILE
testbox: lkp-ne02
tbox_group: lkp-ne02
kconfig: x86_64-rhel
enqueue_time: 2015-03-31 11:57:48.297146032 +08:00
head_commit: 051d101ddcd268a7429d6892c089c1c0858df20b
base_commit: e42391cd048809d903291d07f86ed3934ce138e9
branch: linux-devel/devel-hourly-2015033109
kernel: "/kernel/x86_64-rhel/e42391cd048809d903291d07f86ed3934ce138e9/vmlinuz-4.0.0-rc6"
user: lkp
queue: cyclic
rootfs: debian-x86_64-2015-02-07.cgz
result_root: "/result/lkp-ne02/netperf/performance-300s-200%-TCP_SENDFILE/debian-x86_64-2015-02-07.cgz/x86_64-rhel/e42391cd048809d903291d07f86ed3934ce138e9/0"
node_roles: server client
job_file: "/lkp/scheduled/lkp-ne02/cyclic_netperf-performance-300s-200%-TCP_SENDFILE-x86_64-rhel-BASE-e42391cd048809d903291d07f86ed3934ce138e9-0-20150331-32956-1uhguoe.yaml"
dequeue_time: 2015-03-31 12:29:34.133136287 +08:00
nr_cpu: "$(nproc)"
max_uptime: 1500
modules_initrd: "/kernel/x86_64-rhel/e42391cd048809d903291d07f86ed3934ce138e9/modules.cgz"
bm_initrd: "/lkp/benchmarks/turbostat.cgz,/lkp/benchmarks/netperf-debian.cgz,/lkp/benchmarks/netperf.cgz"
job_state: finished
loadavg: 34.91 29.35 13.49 1/214 9318
start_time: '1427776206'
end_time: '1427776507'
version: "/lkp/lkp/.src-20150331-104755"

echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu10/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu11/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu12/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu13/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu14/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu15/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu2/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu3/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu4/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu5/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu6/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu7/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu8/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu9/cpufreq/scaling_governor netserver
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300
netperf -t TCP_SENDFILE -c -C -l 300

#1 lore raw
Quoted 8 lines · Huang Ying

FYI, we noticed the below changes on

git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
commit b191f9b106ea1a24a711dbebb2925d3313da5852 ("mm: numa: preserve PTE write permissions across a NUMA hinting fault")

testbox/testcase/testparams: lkp-ne02/netperf/performance-300s-200%-TCP_SENDFILE

Quoted 7 lines

bea66fbd11af1ca9 b191f9b106ea1a24a711dbebb2
---------------- --------------------------

         %stddev     %change         %stddev
             \          |                \  
     13115 ±  0%      -8.8%      11968 ±  0%  netperf.time.minor_page_faults
     65.63 ±  6%     -14.6%      56.05 ±  7%  netperf.time.user_time
     65.63 ±  6%     -14.6%      56.05 ±  7%  time.user_time

Hmmm, is this a positive report? It looks like it -- fewer faults at least. Lower user_time is ambiguous because the elapsed time is not included in the report and netperf has variable time depending on whether it can get a confident result or not.

#2 lore raw
Quoted 16 lines · Mel Gorman
Quoted 4 lines · Huang Ying

FYI, we noticed the below changes on

git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master commit b191f9b106ea1a24a711dbebb2925d3313da5852 ("mm: numa: preserve PTE

write permissions across a NUMA hinting fault")

Quoted 3 lines

testbox/testcase/testparams:

lkp-ne02/netperf/performance-300s-200%-TCP_SENDFILE

Quoted 6 lines

bea66fbd11af1ca9 b191f9b106ea1a24a711dbebb2
---------------- --------------------------

         %stddev     %change         %stddev
             \          |                \
     13115 ±  0%      -8.8%      11968 ±  0%
Quoted 6 lines

netperf.time.minor_page_faults

Quoted 2 lines
     65.63 ±  6%     -14.6%      56.05 ±  7%  netperf.time.user_time
     65.63 ±  6%     -14.6%      56.05 ±  7%  time.user_time

Hmmm, is this a positive report? It looks like it -- fewer faults at least.

Yes. I think so too.

Quoted 4 lines

Lower user_time is ambiguous because the elapsed time is not included in the report and netperf has variable time depending on whether it can get a confident result or not.

elapsed time is not included in the report because there is no enough difference between commits, details are as follow:

      2367 ±  0%      +0.2%       2371 ±  0%  time.system_time
    807.50 ±  0%      -0.2%     806.00 ±  0%
time.percent_of_cpu_this_job_got
    301.06 ±  0%      -0.0%     301.05 ±  0%  time.elapsed_time
      1.00 ±  0%      +0.0%       1.00 ±  0%  vmstat.cpu.us
     95.25 ±  1%      +1.8%      97.00 ±  0%  vmstat.cpu.sy
      2.00 ± 50%     -50.0%       1.00 ±  0%  vmstat.cpu.id
      0.00 ±  0%      +NaN%       0.00        vmstat.cpu.wa

Best Regards,
Huang, Ying