From mboxrd@z Thu Jan 1 00:00:00 1970 From: Yuanahn Liu Subject: performance changes on d4b4c2cd: 37.6% fsmark.files_per_sec, -15.9% fsmark.files_per_sec, and few more Date: Wed, 18 Mar 2015 13:03:19 +0800 Message-ID: <20150318050319.GD30894@yliu-dev.sh.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Content-Disposition: inline Sender: linux-kernel-owner@vger.kernel.org To: "shli@kernel.org" Cc: NeilBrown , lkp@01.org, lkp@linux.intel.com, linux-kernel@vger.kernel.org, Jaegeuk Kim , linux-f2fs-devel@lists.sourceforge.net List-Id: linux-f2fs-devel.lists.sourceforge.net Hi, =46YI, we noticed performance changes on `fsmark.files_per_sec' by d4b4= c2cdffab86f5c7594c44635286a6d277d5c6: > commit d4b4c2cdffab86f5c7594c44635286a6d277d5c6 > Author: shli@kernel.org > AuthorDate: Mon Dec 15 12:57:03 2014 +1100 > Commit: NeilBrown > CommitDate: Wed Mar 4 13:40:17 2015 +1100 >=20 > RAID5: batch adjacent full stripe write c1dfe87e41d9c2926fe92f803f02c733ddbccf0b d4b4c2cdffab86f5c7594c4463= 5286a6d277d5c6 ---------------------------------------- --------------------------= -------------- run time(m) metric_value =B1stddev run time(m) metric_v= alue =B1stddev change testbox/benchmark/sub-testcase --- ------ ---------------------------- --- ------ --------------= -------------- -------- ------------------------------ 4 15.3 33.525 =B13.0% 6 11.1 46= =2E133 =B15.0% 37.6% ivb44/fsmark/1x-1t-3HDD-RAID5-xfs-4M-= 120G-NoSync 3 0.5 262.800 =B11.5% 3 0.4 307= =2E367 =B11.2% 17.0% ivb44/fsmark/1x-1t-4BRD_12G-RAID5-f2f= s-4M-30G-NoSync 3 0.5 289.900 =B10.3% 3 0.4 323= =2E367 =B12.4% 11.5% ivb44/fsmark/1x-64t-4BRD_12G-RAID5-f2= fs-4M-30G-NoSync 3 0.5 325.667 =B12.2% 3 0.5 358= =2E800 =B11.8% 10.2% ivb44/fsmark/1x-64t-4BRD_12G-RAID5-ex= t4-4M-30G-NoSync 3 0.6 216.100 =B10.4% 3 0.6 230= =2E100 =B10.4% 6.5% ivb44/fsmark/1x-64t-4BRD_12G-RAID5-f2= fs-4M-30G-fsyncBeforeClose 3 0.5 309.900 =B10.3% 3 0.5 328= =2E500 =B11.1% 6.0% ivb44/fsmark/1x-64t-4BRD_12G-RAID5-xf= s-4M-30G-NoSync 3 13.8 37.000 =B10.2% 3 16.5 31= =2E100 =B10.3% -15.9% ivb44/fsmark/1x-1t-3HDD-RAID5-f2fs-4M= -120G-NoSync NOTE: here are some more info about those test parameters for you to understand the testcase better: 1x : where 'x' means iterations or loop, corresponding to the 'L'= option of fsmark 64t: where 't' means thread 4M : means the single file size, corresponding to the '-s' option= of fsmark 120G, 30G: means the total test size 4BRD_12G: BRD is the ramdisk, where '4' means 4 ramdisk, and wher= e '12G' means the size of one ramdisk. So, it would be 48G in total. = And we made a raid on those ramdisk. And FYI, here I listed more detailed changes for the maximal postive an= d negtive changes. more detailed changes about ivb44/fsmark/1x-1t-3HDD-RAID5-xfs-4M-120G-N= oSync --------- c1dfe87e41d9c292 d4b4c2cdffab86f5c7594c4463 =20 ---------------- -------------------------- =20 %stddev %change %stddev \ | \ =20 33.53 =B1 3% +37.6% 46.13 =B1 4% fsmark.files_per_sec 916 =B1 3% -27.2% 667 =B1 5% fsmark.time.elapsed_t= ime.max 916 =B1 3% -27.2% 667 =B1 5% fsmark.time.elapsed_t= ime 7 =B1 5% +37.6% 10 =B1 6% fsmark.time.percent_o= f_cpu_this_job_got 92097 =B1 2% -23.1% 70865 =B1 4% fsmark.time.voluntary= _context_switches 0.04 =B1 42% +681.0% 0.27 =B1 22% turbostat.Pkg%pc3 716062 =B1 3% -82.7% 124210 =B1 21% cpuidle.C1-IVT.usage 6.883e+08 =B1 2% -86.8% 91146705 =B1 34% cpuidle.C1-IVT.time 0.04 =B1 30% +145.8% 0.10 =B1 25% turbostat.CPU%c3 404 =B1 16% -58.4% 168 =B1 14% cpuidle.POLL.usage 159 =B1 47% +179.5% 444 =B1 23% proc-vmstat.kswapd_lo= w_wmark_hit_quickly 11133 =B1 23% +100.3% 22298 =B1 30% cpuidle.C3-IVT.usage 10286681 =B1 27% +95.6% 20116924 =B1 27% cpuidle.C3-IVT.time 7.92 =B1 16% +77.4% 14.05 =B1 6% turbostat.Pkg%pc6 4.93 =B1 3% -38.6% 3.03 =B1 2% turbostat.CPU%c1 916 =B1 3% -27.2% 667 =B1 5% time.elapsed_time.max 916 =B1 3% -27.2% 667 =B1 5% time.elapsed_time 2137390 =B1 3% -26.7% 1566752 =B1 5% proc-vmstat.pgfault 7 =B1 5% +37.6% 10 =B1 6% time.percent_of_cpu_t= his_job_got 4.309e+10 =B1 3% -26.3% 3.176e+10 =B1 5% cpuidle.C6-IVT.time 49038 =B1 2% -23.9% 37334 =B1 4% uptime.idle 1047 =B1 2% -23.8% 797 =B1 4% uptime.boot 92097 =B1 2% -23.1% 70865 =B1 4% time.voluntary_contex= t_switches 4005888 =B1 0% +13.3% 4537685 =B1 11% meminfo.DirectMap2M 3917 =B1 2% -16.3% 3278 =B1 5% proc-vmstat.pageoutru= n 213737 =B1 1% -13.9% 183969 =B1 3% softirqs.SCHED 46.86 =B1 1% +16.5% 54.59 =B1 1% turbostat.Pkg%pc2 32603 =B1 3% -11.7% 28781 =B1 5% numa-vmstat.node1.nr_= unevictable 130415 =B1 3% -11.7% 115127 =B1 5% numa-meminfo.node1.Un= evictable 256781 =B1 2% -8.8% 234146 =B1 3% softirqs.TASKLET 253606 =B1 2% -8.9% 231108 =B1 3% softirqs.BLOCK 119.10 =B1 2% -70.0% 35.78 =B1 13% iostat.sdc.rrqm/s 119.86 =B1 1% -70.3% 35.64 =B1 12% iostat.sdb.rrqm/s 117.13 =B1 2% -70.2% 34.96 =B1 11% iostat.sda.rrqm/s 504 =B1 2% -67.6% 163 =B1 12% iostat.sdc.rkB/s 507 =B1 1% -67.9% 163 =B1 12% iostat.sdb.rkB/s 496 =B1 2% -67.7% 160 =B1 11% iostat.sda.rkB/s 15392 =B1 3% +37.8% 21203 =B1 5% iostat.sdb.wrqm/s 15393 =B1 3% +37.7% 21203 =B1 5% iostat.sdc.wrqm/s 15392 =B1 3% +37.7% 21203 =B1 5% iostat.sda.wrqm/s 125236 =B1 3% +37.7% 172422 =B1 4% vmstat.io.bo 125181 =B1 3% +37.6% 172303 =B1 4% iostat.md0.wkB/s 552 =B1 3% +37.6% 760 =B1 4% iostat.md0.w/s 62611 =B1 3% +37.6% 86167 =B1 4% iostat.sdb.wkB/s 62613 =B1 3% +37.6% 86167 =B1 4% iostat.sdc.wkB/s 62613 =B1 3% +37.6% 86168 =B1 4% iostat.sda.wkB/s 40.24 =B1 1% -18.5% 32.81 =B1 2% turbostat.CorWatt 200 =B1 0% +22.2% 245 =B1 2% iostat.sdc.w/s 1020 =B1 2% +21.7% 1242 =B1 2% vmstat.system.in 200 =B1 0% +22.1% 245 =B1 2% iostat.sda.w/s 200 =B1 0% +22.2% 245 =B1 2% iostat.sdb.w/s 69.99 =B1 0% -12.4% 61.34 =B1 2% turbostat.PkgWatt 3943 =B1 2% -8.9% 3593 =B1 1% vmstat.system.cs 1.51 =B1 1% +6.1% 1.60 =B1 2% iostat.sdb.avgqu-sz 3.21 =B1 0% +5.4% 3.39 =B1 1% turbostat.RAMWatt 256182 =B1 1% -4.2% 245424 =B1 1% iostat.md0.avgqu-sz more detailed changes about ivb44/fsmark/1x-1t-3HDD-RAID5-f2fs-4M-120G-= NoSync --------- c1dfe87e41d9c292 d4b4c2cdffab86f5c7594c4463 =20 ---------------- -------------------------- =20 %stddev %change %stddev \ | \ =20 37.00 =B1 0% -15.9% 31.10 =B1 0% fsmark.files_per_sec 63414 =B1 4% +57.6% 99945 =B1 1% fsmark.time.voluntary= _context_switches 830 =B1 0% +18.8% 987 =B1 0% fsmark.time.elapsed_t= ime 830 =B1 0% +18.8% 987 =B1 0% fsmark.time.elapsed_t= ime.max 9 =B1 0% -14.8% 7 =B1 6% fsmark.time.percent_o= f_cpu_this_job_got 1.48 =B1 20% +357.3% 6.75 =B1 5% turbostat.Pkg%pc6 63414 =B1 4% +57.6% 99945 =B1 1% time.voluntary_contex= t_switches 109 =B1 15% -37.8% 68 =B1 20% time.involuntary_cont= ext_switches 338 =B1 17% +57.6% 533 =B1 0% cpuidle.POLL.usage 2691 =B1 1% -20.3% 2144 =B1 12% proc-vmstat.kswapd_hi= gh_wmark_hit_quickly 1060792 =B1 0% +20.2% 1275544 =B1 0% cpuidle.C6-IVT.usage 3.876e+10 =B1 0% +19.3% 4.625e+10 =B1 0% cpuidle.C6-IVT.time 830 =B1 0% +18.8% 987 =B1 0% time.elapsed_time.max 830 =B1 0% +18.8% 987 =B1 0% time.elapsed_time 39984 =B1 0% +18.6% 47434 =B1 0% uptime.idle 856 =B1 0% +18.4% 1014 =B1 0% uptime.boot 15874 =B1 12% +20.9% 19188 =B1 6% slabinfo.anon_vma.act= ive_objs 1942445 =B1 0% +18.1% 2293524 =B1 0% proc-vmstat.pgfault 15977 =B1 12% +20.1% 19188 =B1 6% slabinfo.anon_vma.num= _objs 110388 =B1 9% +13.0% 124724 =B1 4% meminfo.DirectMap4k 3107 =B1 8% -20.9% 2459 =B1 15% numa-meminfo.node0.An= onHugePages 18408 =B1 11% +15.0% 21165 =B1 3% slabinfo.free_nid.act= ive_objs 18880 =B1 11% +13.7% 21465 =B1 4% slabinfo.free_nid.num= _objs 1125535 =B1 0% -11.5% 996605 =B1 1% cpuidle.C1-IVT.usage 9 =B1 0% -14.8% 7 =B1 6% time.percent_of_cpu_t= his_job_got 198260 =B1 1% +11.7% 221366 =B1 0% softirqs.SCHED 6.09 =B1 2% -12.2% 5.34 =B1 0% turbostat.CPU%c1 14203 =B1 2% -13.1% 12346 =B1 8% slabinfo.kmalloc-256.= num_objs 13763 =B1 3% -13.3% 11937 =B1 9% slabinfo.kmalloc-256.= active_objs 1255 =B1 6% +10.1% 1383 =B1 1% slabinfo.RAW.num_objs 1255 =B1 6% +10.1% 1383 =B1 1% slabinfo.RAW.active_o= bjs 30.37 =B1 3% +30.5% 39.62 =B1 0% iostat.sdc.rrqm/s 31.23 =B1 5% +28.0% 39.98 =B1 1% iostat.sdb.rrqm/s 33.37 =B1 3% +19.0% 39.72 =B1 2% iostat.sda.rrqm/s 562 =B1 0% -15.9% 472 =B1 0% iostat.md0.w/s 17106 =B1 0% -15.9% 14382 =B1 0% iostat.sda.wrqm/s 17106 =B1 0% -15.9% 14382 =B1 0% iostat.sdc.wrqm/s 17106 =B1 0% -15.9% 14382 =B1 0% iostat.sdb.wrqm/s 69317 =B1 0% -15.9% 58284 =B1 0% iostat.sdc.wkB/s 69316 =B1 0% -15.9% 58284 =B1 0% iostat.sda.wkB/s 69317 =B1 0% -15.9% 58284 =B1 0% iostat.sdb.wkB/s 138603 =B1 0% -15.9% 116543 =B1 0% iostat.md0.wkB/s 138705 =B1 0% -15.9% 116633 =B1 0% vmstat.io.bo 213 =B1 0% -14.5% 182 =B1 0% iostat.sdb.w/s 213 =B1 0% -14.5% 182 =B1 0% iostat.sda.w/s 213 =B1 0% -14.6% 182 =B1 0% iostat.sdc.w/s 4731 =B1 0% -12.7% 4131 =B1 0% vmstat.system.cs 1133 =B1 2% -12.3% 993 =B1 0% vmstat.system.in 3.02 =B1 3% -8.6% 2.76 =B1 3% iostat.sdc.avgqu-sz 3.29 =B1 2% -9.4% 2.98 =B1 3% iostat.sdb.avgqu-sz 25 =B1 19% -21.3% 19 =B1 2% turbostat.Avg_MHz 3.10 =B1 1% -9.4% 2.81 =B1 1% iostat.sda.avgqu-sz 44.45 =B1 1% -5.6% 41.94 =B1 2% turbostat.CorWatt 0.75 =B1 19% -20.1% 0.60 =B1 4% turbostat.%Busy 74.92 =B1 1% -4.9% 71.23 =B1 2% turbostat.PkgWatt