From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga14.intel.com ([192.55.52.115]:17892 "EHLO mga14.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751452AbaHPNKs (ORCPT ); Sat, 16 Aug 2014 09:10:48 -0400 Date: Sat, 16 Aug 2014 21:10:49 +0800 From: Fengguang Wu To: Abhay Sachan Cc: Chris Mason , Dave Hansen , LKML , lkp@01.org, "linux-btrfs@vger.kernel.org" Subject: Re: [btrfs] 4c468fd7485: +7.8% blogbench.write_score, -5.1% turbostat.Pkg_W Message-ID: <20140816131049.GA7157@localhost> References: <20140816075233.GA26842@localhost> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 In-Reply-To: Sender: linux-btrfs-owner@vger.kernel.org List-ID: Hi Abhay, On Sat, Aug 16, 2014 at 05:30:35PM +0530, Abhay Sachan wrote: > Hi Fengguag, > Sorry for the out of topic question, but what benchmark is this? > I have heard about blogbench, but it doesn't give output in this format AFAIK. It is blogbench run in the lkp-tests framework. https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/ It will collect various system stats when blogbench runs. Then it presents you the collected blogbench stats together with slabinfo, meminfo, proc-vmstat, turbostat, softirqs etc. stats. The basic steps to reproduce this report are $ split-job jobs/blogbench.yaml jobs/blogbench.yaml => ./blogbench-1HDD-ext4.yaml jobs/blogbench.yaml => ./blogbench-1HDD-xfs.yaml jobs/blogbench.yaml => ./blogbench-1HDD-btrfs.yaml # requires debian/ubuntu for now $ bin/setup-local --hdd /dev/sdaX ./blogbench-1HDD-btrfs.yaml $ bin/run-local ./blogbench-1HDD-btrfs.yaml The report is generated by the "sbin/compare" script. Thanks, Fengguang > On Sat, Aug 16, 2014 at 1:22 PM, Fengguang Wu wrote: > > Hi Chris, > > > > FYI, we noticed increased performance and reduced power consumption on > > > > commit 4c468fd74859d901c0b78b42bef189295e00d74f ("btrfs: disable strict file flushes for renames and truncates") > > > > test case: lkp-sb02/blogbench/1HDD-btrfs > > > > 0954d74f8f37a47 4c468fd74859d901c0b78b42b > > --------------- ------------------------- > > 1094 ± 1% +7.8% 1180 ± 2% TOTAL blogbench.write_score > > 1396 ±19% -100.0% 0 ± 0% TOTAL slabinfo.btrfs_delalloc_work.active_objs > > 1497 ±17% -100.0% 0 ± 0% TOTAL slabinfo.btrfs_delalloc_work.num_objs > > 426 ±45% -100.0% 0 ± 0% TOTAL proc-vmstat.nr_vmscan_write > > 1.02 ±38% +193.1% 2.99 ±37% TOTAL turbostat.%pc6 > > 0.12 ±48% +113.8% 0.25 ±29% TOTAL turbostat.%pc3 > > 0.38 ±18% +117.7% 0.84 ±34% TOTAL turbostat.%pc2 > > 19377 ±14% -50.9% 9520 ±20% TOTAL proc-vmstat.workingset_refault > > 44 ±41% +68.8% 75 ±28% TOTAL cpuidle.POLL.usage > > 31549 ± 1% +95.7% 61732 ± 1% TOTAL softirqs.BLOCK > > 4547 ±10% -38.3% 2804 ± 9% TOTAL slabinfo.btrfs_ordered_extent.active_objs > > 4628 ±10% -37.1% 2913 ± 9% TOTAL slabinfo.btrfs_ordered_extent.num_objs > > 17597 ± 8% -30.2% 12291 ±14% TOTAL proc-vmstat.nr_writeback > > 70335 ± 8% -30.1% 49174 ±14% TOTAL meminfo.Writeback > > 3606 ± 6% -29.1% 2556 ±10% TOTAL slabinfo.mnt_cache.active_objs > > 14763 ±12% -29.9% 10350 ± 8% TOTAL proc-vmstat.nr_dirty > > 3766 ± 5% -27.8% 2720 ±10% TOTAL slabinfo.mnt_cache.num_objs > > 3509 ± 6% -28.5% 2510 ±11% TOTAL slabinfo.kmalloc-4096.active_objs > > 59201 ±11% -30.1% 41396 ± 8% TOTAL meminfo.Dirty > > 479 ±13% -30.5% 333 ±10% TOTAL slabinfo.kmalloc-4096.num_slabs > > 479 ±13% -30.5% 333 ±10% TOTAL slabinfo.kmalloc-4096.active_slabs > > 3636 ± 6% -26.6% 2669 ±10% TOTAL slabinfo.kmalloc-4096.num_objs > > 6040 ± 8% -28.6% 4314 ± 6% TOTAL slabinfo.kmalloc-96.num_objs > > 5358 ± 5% -25.1% 4011 ± 7% TOTAL slabinfo.kmalloc-96.active_objs > > 757208 ± 4% -22.1% 589874 ± 4% TOTAL meminfo.MemFree > > 189508 ± 4% -22.2% 147518 ± 4% TOTAL proc-vmstat.nr_free_pages > > 762781 ± 4% -21.1% 601525 ± 4% TOTAL vmstat.memory.free > > 10491 ± 2% -16.8% 8725 ± 2% TOTAL slabinfo.kmalloc-64.num_objs > > 2513 ± 4% +16.3% 2923 ± 4% TOTAL slabinfo.kmalloc-128.active_objs > > 9768 ± 3% -15.1% 8298 ± 1% TOTAL slabinfo.kmalloc-64.active_objs > > 2627 ± 4% +14.0% 2995 ± 4% TOTAL slabinfo.kmalloc-128.num_objs > > 96242 ± 2% +15.5% 111120 ± 2% TOTAL slabinfo.btrfs_path.active_objs > > 3448 ± 2% +15.1% 3968 ± 2% TOTAL slabinfo.btrfs_path.num_slabs > > 3448 ± 2% +15.1% 3968 ± 2% TOTAL slabinfo.btrfs_path.active_slabs > > 96580 ± 2% +15.1% 111132 ± 2% TOTAL slabinfo.btrfs_path.num_objs > > 2526 ± 2% +13.5% 2867 ± 1% TOTAL slabinfo.btrfs_extent_state.num_slabs > > 2526 ± 2% +13.5% 2867 ± 1% TOTAL slabinfo.btrfs_extent_state.active_slabs > > 106133 ± 2% +13.5% 120434 ± 1% TOTAL slabinfo.btrfs_extent_state.num_objs > > 104326 ± 2% +12.3% 117189 ± 1% TOTAL slabinfo.btrfs_extent_state.active_objs > > 110759 ± 2% +13.4% 125640 ± 2% TOTAL slabinfo.btrfs_inode.active_objs > > 110759 ± 2% +13.4% 125642 ± 2% TOTAL slabinfo.btrfs_delayed_node.active_objs > > 4261 ± 2% +13.4% 4832 ± 2% TOTAL slabinfo.btrfs_delayed_node.num_slabs > > 4261 ± 2% +13.4% 4832 ± 2% TOTAL slabinfo.btrfs_delayed_node.active_slabs > > 110797 ± 2% +13.4% 125663 ± 2% TOTAL slabinfo.btrfs_delayed_node.num_objs > > 110815 ± 2% +13.4% 125669 ± 2% TOTAL slabinfo.btrfs_inode.num_objs > > 6926 ± 2% +13.4% 7853 ± 2% TOTAL slabinfo.btrfs_inode.num_slabs > > 6926 ± 2% +13.4% 7853 ± 2% TOTAL slabinfo.btrfs_inode.active_slabs > > 5607 ± 3% -11.0% 4991 ± 3% TOTAL slabinfo.kmalloc-256.active_objs > > 6077 ± 2% -9.9% 5476 ± 3% TOTAL slabinfo.kmalloc-256.num_objs > > 11153 ± 1% -7.7% 10295 ± 2% TOTAL proc-vmstat.nr_slab_unreclaimable > > 547824 ± 3% +16.5% 638368 ± 8% TOTAL meminfo.Inactive(file) > > 112124 ± 2% +11.6% 125105 ± 2% TOTAL slabinfo.radix_tree_node.active_objs > > 112169 ± 2% +11.6% 125134 ± 2% TOTAL slabinfo.radix_tree_node.num_objs > > 4005 ± 2% +11.6% 4468 ± 2% TOTAL slabinfo.radix_tree_node.num_slabs > > 4005 ± 2% +11.6% 4468 ± 2% TOTAL slabinfo.radix_tree_node.active_slabs > > 551119 ± 3% +16.4% 641663 ± 8% TOTAL meminfo.Inactive > > 285596 ± 2% +11.4% 318160 ± 2% TOTAL meminfo.SReclaimable > > 156 ± 3% +118.0% 340 ± 2% TOTAL iostat.sda.w/s > > 282 ± 3% -43.2% 160 ± 3% TOTAL iostat.sda.avgrq-sz > > 1.45 ±12% -28.9% 1.03 ±18% TOTAL iostat.sda.rrqm/s > > 633 ± 2% -26.5% 465 ± 2% TOTAL iostat.sda.wrqm/s > > 154423 ± 5% +17.4% 181309 ± 3% TOTAL time.voluntary_context_switches > > 536 ± 5% -11.5% 474 ± 9% TOTAL iostat.sda.await > > 102.71 ± 5% +10.4% 113.36 ± 6% TOTAL iostat.sda.avgqu-sz > > 20842 ± 2% -6.5% 19493 ± 2% TOTAL iostat.sda.wkB/s > > 20856 ± 2% -6.4% 19525 ± 2% TOTAL vmstat.io.bo > > 75.48 ± 4% -6.9% 70.27 ± 5% TOTAL turbostat.%c0 > > 285 ± 4% -6.6% 266 ± 5% TOTAL time.percent_of_cpu_this_job_got > > 34.58 ± 2% -5.5% 32.68 ± 3% TOTAL turbostat.Cor_W > > 39.86 ± 2% -5.1% 37.82 ± 3% TOTAL turbostat.Pkg_W > > 5805 ± 1% -4.3% 5558 ± 3% TOTAL vmstat.system.in > > 10069454 ± 1% +6.3% 10699830 ± 1% TOTAL time.file_system_outputs > > > > > > 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, > > Fengguang > > > > -- > Abhay From mboxrd@z Thu Jan 1 00:00:00 1970 Content-Type: multipart/mixed; boundary="===============8123556881775087877==" MIME-Version: 1.0 From: Fengguang Wu To: lkp@lists.01.org Subject: Re: [btrfs] 4c468fd7485: +7.8% blogbench.write_score, -5.1% turbostat.Pkg_W Date: Sat, 16 Aug 2014 21:10:49 +0800 Message-ID: <20140816131049.GA7157@localhost> In-Reply-To: List-Id: --===============8123556881775087877== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Hi Abhay, On Sat, Aug 16, 2014 at 05:30:35PM +0530, Abhay Sachan wrote: > Hi Fengguag, > Sorry for the out of topic question, but what benchmark is this? > I have heard about blogbench, but it doesn't give output in this format A= FAIK. It is blogbench run in the lkp-tests framework. https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/ It will collect various system stats when blogbench runs. Then it presents you the collected blogbench stats together with slabinfo, meminfo, proc-vmstat, turbostat, softirqs etc. stats. The basic steps to reproduce this report are $ split-job jobs/blogbench.yaml jobs/blogbench.yaml =3D> ./blogbench-1HDD-ext4.yaml jobs/blogbench.yaml =3D> ./blogbench-1HDD-xfs.yaml jobs/blogbench.yaml =3D> ./blogbench-1HDD-btrfs.yaml # requires debian/ubuntu for now $ bin/setup-local --hdd /dev/sdaX ./blogbench-1HDD-btrfs.yaml $ bin/run-local ./blogbench-1HDD-btrfs.yaml The report is generated by the "sbin/compare" script. Thanks, Fengguang > On Sat, Aug 16, 2014 at 1:22 PM, Fengguang Wu = wrote: > > Hi Chris, > > > > FYI, we noticed increased performance and reduced power consumption on > > > > commit 4c468fd74859d901c0b78b42bef189295e00d74f ("btrfs: disable strict= file flushes for renames and truncates") > > > > test case: lkp-sb02/blogbench/1HDD-btrfs > > > > 0954d74f8f37a47 4c468fd74859d901c0b78b42b > > --------------- ------------------------- > > 1094 =C2=B1 1% +7.8% 1180 =C2=B1 2% TOTAL blogbench.w= rite_score > > 1396 =C2=B119% -100.0% 0 =C2=B1 0% TOTAL slabinfo.bt= rfs_delalloc_work.active_objs > > 1497 =C2=B117% -100.0% 0 =C2=B1 0% TOTAL slabinfo.bt= rfs_delalloc_work.num_objs > > 426 =C2=B145% -100.0% 0 =C2=B1 0% TOTAL proc-vmstat= .nr_vmscan_write > > 1.02 =C2=B138% +193.1% 2.99 =C2=B137% TOTAL turbostat.%= pc6 > > 0.12 =C2=B148% +113.8% 0.25 =C2=B129% TOTAL turbostat.%= pc3 > > 0.38 =C2=B118% +117.7% 0.84 =C2=B134% TOTAL turbostat.%= pc2 > > 19377 =C2=B114% -50.9% 9520 =C2=B120% TOTAL proc-vmstat= .workingset_refault > > 44 =C2=B141% +68.8% 75 =C2=B128% TOTAL cpuidle.POL= L.usage > > 31549 =C2=B1 1% +95.7% 61732 =C2=B1 1% TOTAL softirqs.BL= OCK > > 4547 =C2=B110% -38.3% 2804 =C2=B1 9% TOTAL slabinfo.bt= rfs_ordered_extent.active_objs > > 4628 =C2=B110% -37.1% 2913 =C2=B1 9% TOTAL slabinfo.bt= rfs_ordered_extent.num_objs > > 17597 =C2=B1 8% -30.2% 12291 =C2=B114% TOTAL proc-vmstat= .nr_writeback > > 70335 =C2=B1 8% -30.1% 49174 =C2=B114% TOTAL meminfo.Wri= teback > > 3606 =C2=B1 6% -29.1% 2556 =C2=B110% TOTAL slabinfo.mn= t_cache.active_objs > > 14763 =C2=B112% -29.9% 10350 =C2=B1 8% TOTAL proc-vmstat= .nr_dirty > > 3766 =C2=B1 5% -27.8% 2720 =C2=B110% TOTAL slabinfo.mn= t_cache.num_objs > > 3509 =C2=B1 6% -28.5% 2510 =C2=B111% TOTAL slabinfo.km= alloc-4096.active_objs > > 59201 =C2=B111% -30.1% 41396 =C2=B1 8% TOTAL meminfo.Dir= ty > > 479 =C2=B113% -30.5% 333 =C2=B110% TOTAL slabinfo.km= alloc-4096.num_slabs > > 479 =C2=B113% -30.5% 333 =C2=B110% TOTAL slabinfo.km= alloc-4096.active_slabs > > 3636 =C2=B1 6% -26.6% 2669 =C2=B110% TOTAL slabinfo.km= alloc-4096.num_objs > > 6040 =C2=B1 8% -28.6% 4314 =C2=B1 6% TOTAL slabinfo.km= alloc-96.num_objs > > 5358 =C2=B1 5% -25.1% 4011 =C2=B1 7% TOTAL slabinfo.km= alloc-96.active_objs > > 757208 =C2=B1 4% -22.1% 589874 =C2=B1 4% TOTAL meminfo.Mem= Free > > 189508 =C2=B1 4% -22.2% 147518 =C2=B1 4% TOTAL proc-vmstat= .nr_free_pages > > 762781 =C2=B1 4% -21.1% 601525 =C2=B1 4% TOTAL vmstat.memo= ry.free > > 10491 =C2=B1 2% -16.8% 8725 =C2=B1 2% TOTAL slabinfo.km= alloc-64.num_objs > > 2513 =C2=B1 4% +16.3% 2923 =C2=B1 4% TOTAL slabinfo.km= alloc-128.active_objs > > 9768 =C2=B1 3% -15.1% 8298 =C2=B1 1% TOTAL slabinfo.km= alloc-64.active_objs > > 2627 =C2=B1 4% +14.0% 2995 =C2=B1 4% TOTAL slabinfo.km= alloc-128.num_objs > > 96242 =C2=B1 2% +15.5% 111120 =C2=B1 2% TOTAL slabinfo.bt= rfs_path.active_objs > > 3448 =C2=B1 2% +15.1% 3968 =C2=B1 2% TOTAL slabinfo.bt= rfs_path.num_slabs > > 3448 =C2=B1 2% +15.1% 3968 =C2=B1 2% TOTAL slabinfo.bt= rfs_path.active_slabs > > 96580 =C2=B1 2% +15.1% 111132 =C2=B1 2% TOTAL slabinfo.bt= rfs_path.num_objs > > 2526 =C2=B1 2% +13.5% 2867 =C2=B1 1% TOTAL slabinfo.bt= rfs_extent_state.num_slabs > > 2526 =C2=B1 2% +13.5% 2867 =C2=B1 1% TOTAL slabinfo.bt= rfs_extent_state.active_slabs > > 106133 =C2=B1 2% +13.5% 120434 =C2=B1 1% TOTAL slabinfo.bt= rfs_extent_state.num_objs > > 104326 =C2=B1 2% +12.3% 117189 =C2=B1 1% TOTAL slabinfo.bt= rfs_extent_state.active_objs > > 110759 =C2=B1 2% +13.4% 125640 =C2=B1 2% TOTAL slabinfo.bt= rfs_inode.active_objs > > 110759 =C2=B1 2% +13.4% 125642 =C2=B1 2% TOTAL slabinfo.bt= rfs_delayed_node.active_objs > > 4261 =C2=B1 2% +13.4% 4832 =C2=B1 2% TOTAL slabinfo.bt= rfs_delayed_node.num_slabs > > 4261 =C2=B1 2% +13.4% 4832 =C2=B1 2% TOTAL slabinfo.bt= rfs_delayed_node.active_slabs > > 110797 =C2=B1 2% +13.4% 125663 =C2=B1 2% TOTAL slabinfo.bt= rfs_delayed_node.num_objs > > 110815 =C2=B1 2% +13.4% 125669 =C2=B1 2% TOTAL slabinfo.bt= rfs_inode.num_objs > > 6926 =C2=B1 2% +13.4% 7853 =C2=B1 2% TOTAL slabinfo.bt= rfs_inode.num_slabs > > 6926 =C2=B1 2% +13.4% 7853 =C2=B1 2% TOTAL slabinfo.bt= rfs_inode.active_slabs > > 5607 =C2=B1 3% -11.0% 4991 =C2=B1 3% TOTAL slabinfo.km= alloc-256.active_objs > > 6077 =C2=B1 2% -9.9% 5476 =C2=B1 3% TOTAL slabinfo.km= alloc-256.num_objs > > 11153 =C2=B1 1% -7.7% 10295 =C2=B1 2% TOTAL proc-vmstat= .nr_slab_unreclaimable > > 547824 =C2=B1 3% +16.5% 638368 =C2=B1 8% TOTAL meminfo.Ina= ctive(file) > > 112124 =C2=B1 2% +11.6% 125105 =C2=B1 2% TOTAL slabinfo.ra= dix_tree_node.active_objs > > 112169 =C2=B1 2% +11.6% 125134 =C2=B1 2% TOTAL slabinfo.ra= dix_tree_node.num_objs > > 4005 =C2=B1 2% +11.6% 4468 =C2=B1 2% TOTAL slabinfo.ra= dix_tree_node.num_slabs > > 4005 =C2=B1 2% +11.6% 4468 =C2=B1 2% TOTAL slabinfo.ra= dix_tree_node.active_slabs > > 551119 =C2=B1 3% +16.4% 641663 =C2=B1 8% TOTAL meminfo.Ina= ctive > > 285596 =C2=B1 2% +11.4% 318160 =C2=B1 2% TOTAL meminfo.SRe= claimable > > 156 =C2=B1 3% +118.0% 340 =C2=B1 2% TOTAL iostat.sda.= w/s > > 282 =C2=B1 3% -43.2% 160 =C2=B1 3% TOTAL iostat.sda.= avgrq-sz > > 1.45 =C2=B112% -28.9% 1.03 =C2=B118% TOTAL iostat.sda.= rrqm/s > > 633 =C2=B1 2% -26.5% 465 =C2=B1 2% TOTAL iostat.sda.= wrqm/s > > 154423 =C2=B1 5% +17.4% 181309 =C2=B1 3% TOTAL time.volunt= ary_context_switches > > 536 =C2=B1 5% -11.5% 474 =C2=B1 9% TOTAL iostat.sda.= await > > 102.71 =C2=B1 5% +10.4% 113.36 =C2=B1 6% TOTAL iostat.sda.= avgqu-sz > > 20842 =C2=B1 2% -6.5% 19493 =C2=B1 2% TOTAL iostat.sda.= wkB/s > > 20856 =C2=B1 2% -6.4% 19525 =C2=B1 2% TOTAL vmstat.io.bo > > 75.48 =C2=B1 4% -6.9% 70.27 =C2=B1 5% TOTAL turbostat.%= c0 > > 285 =C2=B1 4% -6.6% 266 =C2=B1 5% TOTAL time.percen= t_of_cpu_this_job_got > > 34.58 =C2=B1 2% -5.5% 32.68 =C2=B1 3% TOTAL turbostat.C= or_W > > 39.86 =C2=B1 2% -5.1% 37.82 =C2=B1 3% TOTAL turbostat.P= kg_W > > 5805 =C2=B1 1% -4.3% 5558 =C2=B1 3% TOTAL vmstat.syst= em.in > > 10069454 =C2=B1 1% +6.3% 10699830 =C2=B1 1% TOTAL time.file_s= ystem_outputs > > > > > > Disclaimer: > > Results have been estimated based on internal Intel analysis and are pr= ovided > > for informational purposes only. Any difference in system hardware or s= oftware > > design or configuration may affect actual performance. > > > > Thanks, > > Fengguang > = > = > = > -- = > Abhay --===============8123556881775087877==--