Re: [PATCH-RFC] cfq: Disable low_latency by default for 2.6.32

[Corrado Zoccolo]

On Fri, Nov 27, 2009 at 12:44 PM, Mel Gorman <mel@xxxxxxxxx> wrote:
> On Thu, Nov 26, 2009 at 04:18:18PM +0100, Corrado Zoccolo wrote:
>> > <SNIP>
>> >
>> > In case you mean a partial disabling of cfq_latency, I'm try the
>> > following patch. The intention is to disable the low_latency logic if
>> > kswapd is at work and presumably needs clean pages. Alternative
>> > suggestions welcome.
>
> As it turned out, that patch sucked so I aborted the test and I need to
> think about it a lot more.
What about using the dirty ratio, instead of checking if kswapd is running?

>> Yes, I meant exactly to disable that part, and doing it when kswapd is
>> active is probably a good choice.
>> I have a different idea for 2.6.33, though.
>> If you have a reliable reproducer of the issue, can you test it on
>> git://git.kernel.dk/linux-2.6-block.git branch for-2.6.33?
>> It may already be unaffected, since we had various performance
>> improvements there, but I think a better way to boost writeback is
>> possible.
>>
>
> I haven't tested the high-order allocation scenario yet but the results
> as thing stands are below. There are four kernels being compared
>
> 1. with-low-latency        is 2.6.32-rc8 vanilla
> 2. with-low-latency-block-2.6.33 Âis with the for-2.6.33 from linux-block applied
> 3. with-low-latency-async-rampup Âis with "[RFC,PATCH] cfq-iosched: improve async queue ramp up formula"
> 4. without-low-latency      Âis with low_latency disabled
>
> SYSBENCH
>         sysbench-with    low-latency    low-latency Âsysbench-without
>          low-latency   Âblock-2.6.33   Âasync-rampup    low-latency
> Â Â Â Â Â 1 Â1266.02 ( 0.00%) Â 824.08 (-53.63%) Â1265.15 (-0.07%) Â1278.55 ( 0.98%)
> Â Â Â Â Â 2 Â1182.58 ( 0.00%) Â1226.42 ( 3.57%) Â1223.03 ( 3.31%) Â1379.25 (14.26%)
> Â Â Â Â Â 3 Â1218.64 ( 0.00%) Â1271.38 ( 4.15%) Â1246.42 ( 2.23%) Â1580.08 (22.87%)
> Â Â Â Â Â 4 Â1212.11 ( 0.00%) Â1257.84 ( 3.64%) Â1325.17 ( 8.53%) Â1534.17 (20.99%)
> Â Â Â Â Â 5 Â1046.77 ( 0.00%) Â 981.71 (-6.63%) Â1008.44 (-3.80%) Â1552.48 (32.57%)
> Â Â Â Â Â 6 Â1187.14 ( 0.00%) Â1132.89 (-4.79%) Â1147.18 (-3.48%) Â1661.19 (28.54%)
> Â Â Â Â Â 7 Â1179.37 ( 0.00%) Â1183.61 ( 0.36%) Â1202.49 ( 1.92%) Â 790.26 (-49.24%)
> Â Â Â Â Â 8 Â1164.62 ( 0.00%) Â1143.54 (-1.84%) Â1184.56 ( 1.68%) Â 854.10 (-36.36%)
> Â Â Â Â Â 9 Â1095.22 ( 0.00%) Â1178.72 ( 7.08%) Â1002.42 (-9.26%) Â1655.04 (33.83%)
> Â Â Â Â Â10 Â1147.52 ( 0.00%) Â1153.46 ( 0.52%) Â1151.73 ( 0.37%) Â1653.89 (30.62%)
> Â Â Â Â Â11 Â 823.38 ( 0.00%) Â 820.64 (-0.33%) Â 754.15 (-9.18%) Â1627.45 (49.41%)
> Â Â Â Â Â12 Â 813.73 ( 0.00%) Â 791.44 (-2.82%) Â 848.32 ( 4.08%) Â1494.63 (45.56%)
> Â Â Â Â Â13 Â 898.22 ( 0.00%) Â 789.63 (-13.75%) Â 931.47 ( 3.57%) Â1521.64 (40.97%)
> Â Â Â Â Â14 Â 873.50 ( 0.00%) Â 938.90 ( 6.97%) Â 875.75 ( 0.26%) Â1311.09 (33.38%)
> Â Â Â Â Â15 Â 808.32 ( 0.00%) Â 979.88 (17.51%) Â 877.87 ( 7.92%) Â1009.70 (19.94%)
> Â Â Â Â Â16 Â 758.17 ( 0.00%) Â1096.81 (30.87%) Â 881.23 (13.96%) Â 725.17 (-4.55%)
>
> sysbench is helped by both both block-2.6.33 and async-rampup to some
> extent. For many of the results, plain old disabling low_latency still
> helps the most.
>
> desktop-net-gitk
>           gitk-with    low-latency    low-latency   Âgitk-without
>          low-latency   Âblock-2.6.33   Âasync-rampup    low-latency
> min      Â954.46 ( 0.00%)  570.06 (40.27%)  796.22 (16.58%)  640.65 (32.88%)
> mean      964.79 ( 0.00%)  573.96 (40.51%)  798.01 (17.29%)  655.57 (32.05%)
> stddev     Â10.01 ( 0.00%)   2.65 (73.55%)   1.91 (80.95%)  Â13.33 (-33.18%)
> max      Â981.23 ( 0.00%)  577.21 (41.17%)  800.91 (18.38%)  675.65 (31.14%)
>
> The changes for block in 2.6.33 make a massive difference here, notably
> beating the disabling of low_latency.
Yes. These are read of lots of small files, so the improvements for
seeky workload we introduced in 2.6.33 helps a lot here.
>
> IOZone
>              iozone-with      low-latency      low-latency    Âiozone-without
>              low-latency     Âblock-2.6.33     Âasync-rampup      low-latency
> write-64 Â Â Â Â Â Â Â 151212 ( 0.00%) Â Â Â 163359 ( 7.44%) Â Â Â 163359 ( 7.44%) Â Â Â 159856 ( 5.41%)
> write-128 Â Â Â Â Â Â Â189357 ( 0.00%) Â Â Â 184922 (-2.40%) Â Â Â 202805 ( 6.63%) Â Â Â 206233 ( 8.18%)
> write-256 Â Â Â Â Â Â Â219883 ( 0.00%) Â Â Â 211232 (-4.10%) Â Â Â 189867 (-15.81%) Â Â Â 223174 ( 1.47%)
> write-512 Â Â Â Â Â Â Â224932 ( 0.00%) Â Â Â 222601 (-1.05%) Â Â Â 204459 (-10.01%) Â Â Â 220227 (-2.14%)
> write-1024 Â Â Â Â Â Â 227738 ( 0.00%) Â Â Â 226728 (-0.45%) Â Â Â 216009 (-5.43%) Â Â Â 226155 (-0.70%)
> write-2048 Â Â Â Â Â Â 227564 ( 0.00%) Â Â Â 224167 (-1.52%) Â Â Â 229387 ( 0.79%) Â Â Â 224848 (-1.21%)
> write-4096 Â Â Â Â Â Â 208556 ( 0.00%) Â Â Â 227707 ( 8.41%) Â Â Â 216908 ( 3.85%) Â Â Â 223430 ( 6.66%)
> write-8192 Â Â Â Â Â Â 219484 ( 0.00%) Â Â Â 222365 ( 1.30%) Â Â Â 217737 (-0.80%) Â Â Â 219389 (-0.04%)
> write-16384 Â Â Â Â Â Â206670 ( 0.00%) Â Â Â 209355 ( 1.28%) Â Â Â 204146 (-1.24%) Â Â Â 206295 (-0.18%)
> write-32768 Â Â Â Â Â Â203023 ( 0.00%) Â Â Â 205097 ( 1.01%) Â Â Â 199766 (-1.63%) Â Â Â 201852 (-0.58%)
> write-65536 Â Â Â Â Â Â162134 ( 0.00%) Â Â Â 196670 (17.56%) Â Â Â 189975 (14.66%) Â Â Â 189173 (14.29%)
> write-131072 Â Â Â Â Â Â68534 ( 0.00%) Â Â Â Â69145 ( 0.88%) Â Â Â Â64519 (-6.22%) Â Â Â Â67417 (-1.66%)
> write-262144 Â Â Â Â Â Â32936 ( 0.00%) Â Â Â Â28587 (-15.21%) Â Â Â Â31470 (-4.66%) Â Â Â Â27750 (-18.69%)
> write-524288 Â Â Â Â Â Â24044 ( 0.00%) Â Â Â Â23560 (-2.05%) Â Â Â Â23116 (-4.01%) Â Â Â Â23759 (-1.20%)
> rewrite-64 Â Â Â Â Â Â 755681 ( 0.00%) Â Â Â 800767 ( 5.63%) Â Â Â 469931 (-60.81%) Â Â Â 755681 ( 0.00%)
> rewrite-128 Â Â Â Â Â Â581518 ( 0.00%) Â Â Â 639723 ( 9.10%) Â Â Â 591774 ( 1.73%) Â Â Â 799840 (27.30%)
> rewrite-256 Â Â Â Â Â Â639427 ( 0.00%) Â Â Â 710511 (10.00%) Â Â Â 666414 ( 4.05%) Â Â Â 659861 ( 3.10%)
> rewrite-512 Â Â Â Â Â Â669577 ( 0.00%) Â Â Â 743788 ( 9.98%) Â Â Â 692017 ( 3.24%) Â Â Â 684954 ( 2.24%)
> rewrite-1024 Â Â Â Â Â 680960 ( 0.00%) Â Â Â 755195 ( 9.83%) Â Â Â 701422 ( 2.92%) Â Â Â 686182 ( 0.76%)
> rewrite-2048 Â Â Â Â Â 685263 ( 0.00%) Â Â Â 743123 ( 7.79%) Â Â Â 703445 ( 2.58%) Â Â Â 692780 ( 1.09%)
> rewrite-4096 Â Â Â Â Â 631352 ( 0.00%) Â Â Â 686776 ( 8.07%) Â Â Â 640007 ( 1.35%) Â Â Â 643266 ( 1.85%)
> rewrite-8192 Â Â Â Â Â 442146 ( 0.00%) Â Â Â 474089 ( 6.74%) Â Â Â 457768 ( 3.41%) Â Â Â 442624 ( 0.11%)
> rewrite-16384 Â Â Â Â Â428641 ( 0.00%) Â Â Â 454857 ( 5.76%) Â Â Â 442896 ( 3.22%) Â Â Â 432613 ( 0.92%)
> rewrite-32768 Â Â Â Â Â425361 ( 0.00%) Â Â Â 444206 ( 4.24%) Â Â Â 434472 ( 2.10%) Â Â Â 430568 ( 1.21%)
> rewrite-65536 Â Â Â Â Â405183 ( 0.00%) Â Â Â 433898 ( 6.62%) Â Â Â 419843 ( 3.49%) Â Â Â 389242 (-4.10%)
> rewrite-131072 Â Â Â Â Â66110 ( 0.00%) Â Â Â Â58370 (-13.26%) Â Â Â Â54342 (-21.66%) Â Â Â Â58472 (-13.06%)
> rewrite-262144 Â Â Â Â Â29254 ( 0.00%) Â Â Â Â24665 (-18.61%) Â Â Â Â25710 (-13.78%) Â Â Â Â29306 ( 0.18%)
> rewrite-524288 Â Â Â Â Â23812 ( 0.00%) Â Â Â Â20742 (-14.80%) Â Â Â Â22490 (-5.88%) Â Â Â Â24543 ( 2.98%)
> read-64 Â Â Â Â Â Â Â Â934589 ( 0.00%) Â Â Â1160938 (19.50%) Â Â Â1004538 ( 6.96%) Â Â Â 840903 (-11.14%)
> read-128 Â Â Â Â Â Â Â1601534 ( 0.00%) Â Â Â1869179 (14.32%) Â Â Â1681806 ( 4.77%) Â Â Â1280633 (-25.06%)
> read-256 Â Â Â Â Â Â Â1255511 ( 0.00%) Â Â Â1526887 (17.77%) Â Â Â1304314 ( 3.74%) Â Â Â1310683 ( 4.21%)
> read-512 Â Â Â Â Â Â Â1291158 ( 0.00%) Â Â Â1377278 ( 6.25%) Â Â Â1336145 ( 3.37%) Â Â Â1319723 ( 2.16%)
> read-1024 Â Â Â Â Â Â 1319408 ( 0.00%) Â Â Â1306564 (-0.98%) Â Â Â1368162 ( 3.56%) Â Â Â1347557 ( 2.09%)
> read-2048 Â Â Â Â Â Â 1316016 ( 0.00%) Â Â Â1394645 ( 5.64%) Â Â Â1339827 ( 1.78%) Â Â Â1347393 ( 2.33%)
> read-4096 Â Â Â Â Â Â 1253710 ( 0.00%) Â Â Â1307525 ( 4.12%) Â Â Â1247519 (-0.50%) Â Â Â1251882 (-0.15%)
> read-8192 Â Â Â Â Â Â Â995149 ( 0.00%) Â Â Â1033337 ( 3.70%) Â Â Â1016944 ( 2.14%) Â Â Â1011794 ( 1.65%)
> read-16384 Â Â Â Â Â Â 883156 ( 0.00%) Â Â Â 905213 ( 2.44%) Â Â Â 905213 ( 2.44%) Â Â Â 897458 ( 1.59%)
> read-32768 Â Â Â Â Â Â 844368 ( 0.00%) Â Â Â 855213 ( 1.27%) Â Â Â 849609 ( 0.62%) Â Â Â 856364 ( 1.40%)
> read-65536 Â Â Â Â Â Â 816099 ( 0.00%) Â Â Â 839262 ( 2.76%) Â Â Â 835019 ( 2.27%) Â Â Â 826473 ( 1.26%)
> read-131072 Â Â Â Â Â Â818055 ( 0.00%) Â Â Â 837369 ( 2.31%) Â Â Â 828230 ( 1.23%) Â Â Â 824351 ( 0.76%)
> read-262144 Â Â Â Â Â Â827225 ( 0.00%) Â Â Â 839635 ( 1.48%) Â Â Â 840538 ( 1.58%) Â Â Â 835693 ( 1.01%)
> read-524288 Â Â Â Â Â Â 24653 ( 0.00%) Â Â Â Â21387 (-15.27%) Â Â Â Â20602 (-19.66%) Â Â Â Â22519 (-9.48%)
> reread-64 Â Â Â Â Â Â 2329708 ( 0.00%) Â Â Â2251544 (-3.47%) Â Â Â1985134 (-17.36%) Â Â Â1985134 (-17.36%)
> reread-128 Â Â Â Â Â Â1446222 ( 0.00%) Â Â Â1979446 (26.94%) Â Â Â2009076 (28.02%) Â Â Â2137031 (32.33%)
> reread-256 Â Â Â Â Â Â1828508 ( 0.00%) Â Â Â2006158 ( 8.86%) Â Â Â1892980 ( 3.41%) Â Â Â1879725 ( 2.72%)
> reread-512 Â Â Â Â Â Â1521718 ( 0.00%) Â Â Â1642783 ( 7.37%) Â Â Â1508887 (-0.85%) Â Â Â1579934 ( 3.68%)
> reread-1024 Â Â Â Â Â 1347557 ( 0.00%) Â Â Â1422540 ( 5.27%) Â Â Â1384034 ( 2.64%) Â Â Â1375171 ( 2.01%)
> reread-2048 Â Â Â Â Â 1340664 ( 0.00%) Â Â Â1413929 ( 5.18%) Â Â Â1372364 ( 2.31%) Â Â Â1350783 ( 0.75%)
> reread-4096 Â Â Â Â Â 1259592 ( 0.00%) Â Â Â1324868 ( 4.93%) Â Â Â1273788 ( 1.11%) Â Â Â1284839 ( 1.96%)
> reread-8192 Â Â Â Â Â 1007285 ( 0.00%) Â Â Â1033710 ( 2.56%) Â Â Â1027159 ( 1.93%) Â Â Â1011317 ( 0.40%)
> reread-16384 Â Â Â Â Â 891404 ( 0.00%) Â Â Â 910828 ( 2.13%) Â Â Â 916562 ( 2.74%) Â Â Â 905022 ( 1.50%)
> reread-32768 Â Â Â Â Â 850492 ( 0.00%) Â Â Â 859341 ( 1.03%) Â Â Â 856385 ( 0.69%) Â Â Â 862772 ( 1.42%)
> reread-65536 Â Â Â Â Â 836565 ( 0.00%) Â Â Â 852664 ( 1.89%) Â Â Â 852315 ( 1.85%) Â Â Â 847020 ( 1.23%)
> reread-131072 Â Â Â Â Â844516 ( 0.00%) Â Â Â 862590 ( 2.10%) Â Â Â 854067 ( 1.12%) Â Â Â 853155 ( 1.01%)
> reread-262144 Â Â Â Â Â851524 ( 0.00%) Â Â Â 860559 ( 1.05%) Â Â Â 864921 ( 1.55%) Â Â Â 860653 ( 1.06%)
> reread-524288 Â Â Â Â Â 24927 ( 0.00%) Â Â Â Â21300 (-17.03%) Â Â Â Â19748 (-26.23%) Â Â Â Â22487 (-10.85%)
> randread-64 Â Â Â Â Â 1605256 ( 0.00%) Â Â Â1605256 ( 0.00%) Â Â Â1605256 ( 0.00%) Â Â Â1775099 ( 9.57%)
> randread-128 Â Â Â Â Â1179358 ( 0.00%) Â Â Â1582649 (25.48%) Â Â Â1511363 (21.97%) Â Â Â1528576 (22.85%)
> randread-256 Â Â Â Â Â1421755 ( 0.00%) Â Â Â1599680 (11.12%) Â Â Â1460430 ( 2.65%) Â Â Â1310683 (-8.47%)
> randread-512 Â Â Â Â Â1306873 ( 0.00%) Â Â Â1278855 (-2.19%) Â Â Â1243315 (-5.11%) Â Â Â1281909 (-1.95%)
> randread-1024 Â Â Â Â 1201314 ( 0.00%) Â Â Â1254656 ( 4.25%) Â Â Â1190657 (-0.90%) Â Â Â1231629 ( 2.46%)
> randread-2048 Â Â Â Â 1179413 ( 0.00%) Â Â Â1227971 ( 3.95%) Â Â Â1185272 ( 0.49%) Â Â Â1190529 ( 0.93%)
> randread-4096 Â Â Â Â 1107005 ( 0.00%) Â Â Â1160862 ( 4.64%) Â Â Â1110727 ( 0.34%) Â Â Â1116792 ( 0.88%)
> randread-8192 Â Â Â Â Â894337 ( 0.00%) Â Â Â 924264 ( 3.24%) Â Â Â 912676 ( 2.01%) Â Â Â 899487 ( 0.57%)
> randread-16384 Â Â Â Â 783760 ( 0.00%) Â Â Â 800299 ( 2.07%) Â Â Â 793351 ( 1.21%) Â Â Â 791341 ( 0.96%)
> randread-32768 Â Â Â Â 740498 ( 0.00%) Â Â Â 743720 ( 0.43%) Â Â Â 741233 ( 0.10%) Â Â Â 743511 ( 0.41%)
> randread-65536 Â Â Â Â 721640 ( 0.00%) Â Â Â 727692 ( 0.83%) Â Â Â 726984 ( 0.74%) Â Â Â 728139 ( 0.89%)
> randread-131072 Â Â Â Â715284 ( 0.00%) Â Â Â 722094 ( 0.94%) Â Â Â 717746 ( 0.34%) Â Â Â 720825 ( 0.77%)
> randread-262144 Â Â Â Â709855 ( 0.00%) Â Â Â 706770 (-0.44%) Â Â Â 709133 (-0.10%) Â Â Â 714943 ( 0.71%)
> randread-524288 Â Â Â Â Â 394 ( 0.00%) Â Â Â Â Â421 ( 6.41%) Â Â Â Â Â418 ( 5.74%) Â Â Â Â Â431 ( 8.58%)
> randwrite-64 Â Â Â Â Â 730988 ( 0.00%) Â Â Â 764288 ( 4.36%) Â Â Â 723111 (-1.09%) Â Â Â 730988 ( 0.00%)
> randwrite-128 Â Â Â Â Â746459 ( 0.00%) Â Â Â 799840 ( 6.67%) Â Â Â 746459 ( 0.00%) Â Â Â 742331 (-0.56%)
> randwrite-256 Â Â Â Â Â695778 ( 0.00%) Â Â Â 752329 ( 7.52%) Â Â Â 720041 ( 3.37%) Â Â Â 727850 ( 4.41%)
> randwrite-512 Â Â Â Â Â666253 ( 0.00%) Â Â Â 722760 ( 7.82%) Â Â Â 667081 ( 0.12%) Â Â Â 691126 ( 3.60%)
> randwrite-1024 Â Â Â Â 651223 ( 0.00%) Â Â Â 697776 ( 6.67%) Â Â Â 663292 ( 1.82%) Â Â Â 659625 ( 1.27%)
> randwrite-2048 Â Â Â Â 655558 ( 0.00%) Â Â Â 691887 ( 5.25%) Â Â Â 665720 ( 1.53%) Â Â Â 664073 ( 1.28%)
> randwrite-4096 Â Â Â Â 635556 ( 0.00%) Â Â Â 662721 ( 4.10%) Â Â Â 643170 ( 1.18%) Â Â Â 642400 ( 1.07%)
> randwrite-8192 Â Â Â Â 467357 ( 0.00%) Â Â Â 491364 ( 4.89%) Â Â Â 476720 ( 1.96%) Â Â Â 469734 ( 0.51%)
> randwrite-16384 Â Â Â Â413188 ( 0.00%) Â Â Â 427521 ( 3.35%) Â Â Â 417353 ( 1.00%) Â Â Â 417282 ( 0.98%)
> randwrite-32768 Â Â Â Â404161 ( 0.00%) Â Â Â 411721 ( 1.84%) Â Â Â 404942 ( 0.19%) Â Â Â 407580 ( 0.84%)
> randwrite-65536 Â Â Â Â379372 ( 0.00%) Â Â Â 397312 ( 4.52%) Â Â Â 386853 ( 1.93%) Â Â Â 381273 ( 0.50%)
> randwrite-131072 Â Â Â Â21780 ( 0.00%) Â Â Â Â16924 (-28.69%) Â Â Â Â21177 (-2.85%) Â Â Â Â19758 (-10.23%)
> randwrite-262144 Â Â Â Â 6249 ( 0.00%) Â Â Â Â 5548 (-12.64%) Â Â Â Â 6370 ( 1.90%) Â Â Â Â 6316 ( 1.06%)
> randwrite-524288 Â Â Â Â 2915 ( 0.00%) Â Â Â Â 2582 (-12.90%) Â Â Â Â 2871 (-1.53%) Â Â Â Â 2859 (-1.96%)
> bkwdread-64 Â Â Â Â Â 1141196 ( 0.00%) Â Â Â1141196 ( 0.00%) Â Â Â1004538 (-13.60%) Â Â Â1141196 ( 0.00%)
> bkwdread-128 Â Â Â Â Â1066865 ( 0.00%) Â Â Â1386465 (23.05%) Â Â Â1400936 (23.85%) Â Â Â1101900 ( 3.18%)
> bkwdread-256 Â Â Â Â Â 877797 ( 0.00%) Â Â Â1105556 (20.60%) Â Â Â1105556 (20.60%) Â Â Â1105556 (20.60%)
> bkwdread-512 Â Â Â Â Â1133103 ( 0.00%) Â Â Â1162547 ( 2.53%) Â Â Â1175271 ( 3.59%) Â Â Â1162547 ( 2.53%)
> bkwdread-1024 Â Â Â Â 1163562 ( 0.00%) Â Â Â1206714 ( 3.58%) Â Â Â1213534 ( 4.12%) Â Â Â1195962 ( 2.71%)
> bkwdread-2048 Â Â Â Â 1163439 ( 0.00%) Â Â Â1218910 ( 4.55%) Â Â Â1204552 ( 3.41%) Â Â Â1204552 ( 3.41%)
> bkwdread-4096 Â Â Â Â 1116792 ( 0.00%) Â Â Â1175477 ( 4.99%) Â Â Â1159922 ( 3.72%) Â Â Â1150600 ( 2.94%)
> bkwdread-8192 Â Â Â Â Â912288 ( 0.00%) Â Â Â 935233 ( 2.45%) Â Â Â 944695 ( 3.43%) Â Â Â 934724 ( 2.40%)
> bkwdread-16384 Â Â Â Â 817707 ( 0.00%) Â Â Â 824140 ( 0.78%) Â Â Â 832527 ( 1.78%) Â Â Â 829152 ( 1.38%)
> bkwdread-32768 Â Â Â Â 775898 ( 0.00%) Â Â Â 773714 (-0.28%) Â Â Â 785494 ( 1.22%) Â Â Â 787691 ( 1.50%)
> bkwdread-65536 Â Â Â Â 759643 ( 0.00%) Â Â Â 769924 ( 1.34%) Â Â Â 778780 ( 2.46%) Â Â Â 772174 ( 1.62%)
> bkwdread-131072 Â Â Â Â763215 ( 0.00%) Â Â Â 769634 ( 0.83%) Â Â Â 773707 ( 1.36%) Â Â Â 773816 ( 1.37%)
> bkwdread-262144 Â Â Â Â765491 ( 0.00%) Â Â Â 768992 ( 0.46%) Â Â Â 780876 ( 1.97%) Â Â Â 780021 ( 1.86%)
> bkwdread-524288 Â Â Â Â Â3688 ( 0.00%) Â Â Â Â 3595 (-2.59%) Â Â Â Â 3577 (-3.10%) Â Â Â Â 3724 ( 0.97%)
>
> The upcoming changes for 2.6.33 also help iozone in many cases, often by more
> than just disabling low_latency. It has the occasional massive gain or loss
> for the larger file sizes. I don't know why this is but as the big losses
> appear to be mostly in the write-tests, I would guess that it's differences
> in heavy-writer-throttling.
I wonder if 2.6.33 + my async rampup patch will improve still further,
maybe reaching the low_latency=0 performance also for writing tests.
>
> The only downside with block-2.6.33 is that there are a lot of patches in
> there and doesn't help with the 2.6.32 release as such. I could do a reverse
> bisect to see what helps the most in there but under ideal conditions, it'll
> take 3 days to complete and I wouldn't be able to start until Monday as I'm
> out of the country for the weekend. That's a bit late.
Bisect will likely not help, since we have several patch series with
heavy internal dependencies in that tree.
If one of the patch series is found to bring the improvement, you have
to backport the entire series, that is not advisable for a rc8 or for
stable.
>
> p.s. As a consequence of being out of the country, I also won't be able to
> Â Â respond to mail over the weekend.
>
> --
> Mel Gorman
>
Thanks for the detailed report
Corrado
¢éì®&Þ~º&¶¬–+-±éÝ¥Šw®žË±Êâmébžìdz¹Þ)í…æèw*jg¬±¨¶‰šŽŠÝj/êäz¹ÞŠà2ŠÞ¨è­Ú&¢)ß«a¶Úþø®G«éh®æj:+v‰¨Šwè†Ù>Wš±êÞiÛaxPjØm¶ŸÿÃ-»+ƒùdš_