* [benchmarks] on recent kernels
@ 2003-04-01 23:29 rwhron
2003-04-03 14:07 ` Andrea Arcangeli
0 siblings, 1 reply; 2+ messages in thread
From: rwhron @ 2003-04-01 23:29 UTC (permalink / raw)
To: linux-kernel
Post for posterity. This is on a uniprocessor K6/2 with 384
MB ram and 2 IDE disks. Generally only one disk at a time
is used for the tests below.
AIM7 is a mixture of very small jobs from the AIM9 benchmark.
The Tasks column is the simulated load average.
A copy of the workfile for the custom load is at:
http://home.earthlink.net/~rwhron/kernel/workfile.custom
There isn't any notion of fairness here, just speed.
AIM7 custom workload
kernel Tasks Jobs/Min Real CPU
2.4.21-pre5-jam1 8 535.7 95.9 85.9 seconds
2.4.21-pre5aa2 8 526.9 97.5 87.5
2.4.21-pre5 8 523.3 98.2 88.1
2.4.21-pre5-ac3 8 513.9 99.9 89.9
2.5.66-mjb1 8 486.0 105.7 95.0
2.5.66 8 470.8 109.1 98.3
2.5.66-mm1 8 460.8 111.5 100.5
2.5.66-ac1 8 455.7 112.7 101.7
2.4.21-pre5aa2 104 787.0 848.4 838.1
2.4.21-pre5-ac3 104 762.1 876.1 865.8
2.4.21-pre5-jam1 104 759.7 878.8 868.5
2.4.21-pre5 104 757.0 882.0 871.7
2.5.66-mjb1 104 743.4 898.1 888.1
2.5.66 104 728.8 916.1 905.7
2.5.66-ac1 104 719.1 928.5 918.0
2.5.66-mm1 104 708.7 942.1 931.7
2.4.21-pre5aa2 200 796.2 1612.7 1601.5
2.4.21-pre5-jam1 200 779.7 1646.8 1635.4
2.4.21-pre5-ac3 200 775.7 1655.2 1643.9
2.4.21-pre5 200 771.3 1664.8 1653.6
2.5.66-mjb1 200 760.3 1688.8 1678.8
2.5.66 200 746.3 1720.5 1709.1
2.5.66-ac1 200 738.6 1738.3 1727.1
2.5.66-mm1 200 728.8 1761.7 1750.6
2.4.21-pre5aa2 296 806.2 2357.1 2344.4
2.4.21-pre5-ac3 296 790.1 2405.3 2392.5
2.4.21-pre5-jam1 296 789.9 2405.9 2392.8
2.4.21-pre5 296 775.5 2450.4 2437.9
2.5.66-mjb1 296 764.3 2486.2 2475.9
2.5.66 296 754.1 2519.9 2507.5
2.5.66-ac1 296 748.2 2540.0 2527.3
2.5.66-mm1 296 733.1 2592.1 2579.5
2.4.21-pre5aa2 392 791.8 3178.2 3159.6
2.4.21-pre5-ac3 392 783.8 3210.7 3192.9
2.4.21-pre5-jam1 392 777.8 3235.8 3214.7
2.5.66-mjb1 392 770.7 3265.5 3255.0
2.4.21-pre5 392 766.2 3284.5 3269.2
2.5.66 392 753.8 3338.5 3325.1
2.5.66-ac1 392 748.9 3360.7 3347.2
2.5.66-mm1 392 738.5 3407.6 3393.7
AIM7 dbase workload
This load has more synchronous writes. Supposed to be a little
like a database.
2.4.21-pre5-akpm has the -aa patches akpm posted for 2.4.21-pre4.
kernel Tasks Jobs/Min Real CPU
2.4.21-pre5-akpm 4 43.0 552.7 118.2 seconds
2.4.21-pre5 4 42.9 554.1 115.4
2.4.21-pre5aa2 4 42.8 554.9 116.6
2.4.21-pre5-ac3 4 42.7 555.9 117.5
2.5.66 4 42.0 565.3 124.9
2.5.66-mm1 4 41.9 566.6 126.0
2.4.21-pre5-jam1 4 41.8 568.9 115.0
2.5.66-ac1 4 41.2 577.1 125.0
2.5.66-mjb1 4 40.7 583.1 122.9
2.2.23 4 40.2 590.9 110.0
2.4.21-pre5-akpm 8 75.0 633.8 204.7
2.4.21-pre5aa2 8 74.7 636.4 199.9
2.4.21-pre5 8 74.3 639.9 201.6
2.4.21-pre5-ac3 8 74.0 642.0 201.9
2.2.23 8 70.1 677.8 198.1
2.5.66-mjb1 8 69.1 687.4 208.5
2.5.66 8 68.7 691.4 209.6
2.5.66-ac1 8 68.7 692.1 209.1
2.5.66-mm1 8 68.5 693.7 213.6
2.4.21-pre5-jam1 8 68.1 697.8 200.2
2.4.21-pre5aa2 12 97.7 729.8 284.9
2.4.21-pre5-akpm 12 97.6 730.0 287.5
2.4.21-pre5 12 96.9 735.3 286.3
2.4.21-pre5-ac3 12 95.4 746.8 287.1
2.2.23 12 91.0 783.0 283.3
2.5.66-mjb1 12 86.3 826.0 293.7
2.5.66 12 85.9 829.4 298.4
2.4.21-pre5-jam1 12 85.4 834.4 284.9
2.5.66-mm1 12 85.3 835.5 300.1
2.5.66-ac1 12 84.3 845.2 296.2
2.4.21-pre5 16 116.1 818.9 376.7
2.4.21-pre5-akpm 16 115.9 820.2 376.1
2.4.21-pre5-ac3 16 115.8 820.9 372.9
2.4.21-pre5aa2 16 115.3 824.0 366.8
2.2.23 16 111.2 854.6 385.7
2.5.66 16 99.3 957.1 379.9
2.5.66-ac1 16 98.8 961.6 387.1
2.5.66-mjb1 16 98.5 964.6 374.3
2.4.21-pre5-jam1 16 98.1 969.2 365.2
2.5.66-mm1 16 97.6 973.6 388.3
2.4.21-pre5-akpm 20 132.2 898.5 464.4
2.4.21-pre5-ac3 20 130.4 910.9 456.2
2.4.21-pre5 20 128.6 923.7 465.7
2.4.21-pre5aa2 20 124.8 952.1 450.7
2.2.23 20 122.2 972.1 470.0
2.5.66-mm1 20 112.5 1055.6 476.6
2.5.66 20 110.8 1072.0 467.1
2.5.66-mjb1 20 110.8 1072.1 459.1
2.5.66-ac1 20 109.3 1086.7 468.4
2.4.21-pre5-jam1 20 108.5 1094.9 448.7
2.4.21-pre5aa2 24 145.0 983.2 533.3
2.4.21-pre5-akpm 24 144.7 985.2 559.4
2.4.21-pre5-ac3 24 144.0 990.2 541.3
2.4.21-pre5 24 143.6 992.7 552.9
2.2.23 24 138.0 1033.2 569.2
2.5.66 24 127.3 1119.5 552.1
2.5.66-mjb1 24 126.6 1126.4 549.0
2.4.21-pre5-jam1 24 126.1 1130.1 536.1
2.5.66-mm1 24 125.9 1131.9 562.4
2.5.66-ac1 24 125.8 1133.0 559.8
2.4.21-pre5-akpm 28 151.4 1098.2 644.1
2.4.21-pre5aa2 28 150.6 1104.2 617.6
2.4.21-pre5-ac3 28 150.2 1107.2 624.9
2.4.21-pre5 28 150.1 1108.1 647.4
2.2.23 28 141.5 1175.7 664.2
2.5.66-mm1 28 131.9 1261.3 648.5
2.5.66-ac1 28 130.0 1279.1 644.6
2.5.66 28 129.7 1282.4 641.6
2.5.66-mjb1 28 129.7 1282.2 634.0
2.4.21-pre5-jam1 28 129.6 1282.9 619.4
2.4.21-pre5-ac3 32 157.5 1207.0 711.3
2.4.21-pre5 32 156.1 1218.0 738.6
2.4.21-pre5-akpm 32 155.8 1220.3 738.5
2.4.21-pre5aa2 32 153.6 1237.6 698.6
2.2.23 32 151.5 1254.8 764.8
2.5.66-mm1 32 141.3 1345.2 736.1
2.5.66 32 139.6 1361.7 726.6
2.5.66-ac1 32 139.5 1362.2 725.0
2.5.66-mjb1 32 138.9 1368.7 717.6
2.4.21-pre5-jam1 32 138.6 1371.7 704.0
AIM7 fserver workload
kernel Tasks Jobs/Min Real CPU
2.4.21-pre5-jam1 1 53.9 112.4 22.6
2.4.21-pre5aa2 1 52.4 115.7 23.1
2.4.21-pre5-akpm 1 51.1 118.7 24.0
2.4.21-pre5 1 50.7 119.6 23.6
2.4.21-pre5-ac3 1 50.3 120.5 26.6
2.5.66-mjb1 1 48.3 125.4 26.0
2.5.66 1 47.7 127.0 27.9
2.5.66-ac1 1 46.4 130.5 28.7
2.2.23 1 45.6 132.8 46.4
2.5.66-mm1 1 45.4 133.5 28.0
2.4.21-pre5 2 94.0 129.0 44.7
2.4.21-pre5-akpm 2 93.7 129.4 46.1
2.2.23 2 91.6 132.3 45.6
2.5.66-ac1 2 84.9 142.8 51.5
2.4.21-pre5-jam1 2 83.4 145.2 40.6
2.4.21-pre5aa2 2 83.4 145.3 45.1
2.5.66 2 83.1 145.8 50.9
2.5.66-mm1 2 82.9 146.2 49.3
2.4.21-pre5-ac3 2 82.8 146.3 49.2
2.5.66-mjb1 2 81.2 149.3 51.0
2.4.21-pre5 3 122.0 149.0 57.9
2.4.21-pre5-jam1 3 117.5 154.8 57.4
2.4.21-pre5aa2 3 117.2 155.1 57.9
2.4.21-pre5-akpm 3 115.2 157.9 58.3
2.4.21-pre5-ac3 3 114.2 159.2 63.3
2.5.66 3 111.5 163.0 69.1
2.5.66-ac1 3 110.9 163.9 71.4
2.2.23 3 110.2 164.9 56.4
2.5.66-mm1 3 110.1 165.2 70.0
2.5.66-mjb1 3 103.6 175.4 66.1
2.4.21-pre5-ac3 4 134.1 180.8 77.4
2.4.21-pre5-jam1 4 132.7 182.7 71.1
2.4.21-pre5 4 131.6 184.2 73.9
2.4.21-pre5aa2 4 128.4 188.7 77.3
2.4.21-pre5-akpm 4 125.9 192.6 72.5
2.5.66-mm1 4 125.7 192.8 84.4
2.5.66 4 121.1 200.2 85.1
2.5.66-mjb1 4 117.4 206.5 85.1
2.2.23 4 115.6 209.7 71.7
2.5.66-ac1 4 115.0 210.7 90.4
2.4.21-pre5 5 143.4 211.4 85.8
2.4.21-pre5-ac3 5 142.1 213.2 94.2
2.4.21-pre5-akpm 5 137.9 219.7 86.0
2.2.23 5 137.6 220.3 95.6
2.4.21-pre5-jam1 5 133.3 227.3 86.0
2.4.21-pre5aa2 5 131.5 230.5 87.7
2.5.66 5 131.4 230.6 105.7
2.5.66-mjb1 5 130.4 232.3 101.6
2.5.66-ac1 5 127.4 237.8 104.4
2.5.66-mm1 5 125.7 241.0 106.4
2.4.21-pre5 6 158.2 229.9 96.9
2.4.21-pre5-ac3 6 157.5 230.8 107.1
2.4.21-pre5-akpm 6 155.4 234.0 99.6
2.4.21-pre5aa2 6 150.5 241.5 103.9
2.4.21-pre5-jam1 6 148.7 244.5 97.8
2.5.66-ac1 6 144.7 251.3 122.0
2.5.66-mjb1 6 144.4 251.8 120.1
2.5.66 6 138.1 263.3 115.2
2.2.23 6 137.4 264.7 117.9
2.5.66-mm1 6 133.9 271.6 121.2
2.4.21-pre5aa2 7 142.1 298.5 114.5
2.4.21-pre5 7 141.0 300.8 110.6
2.4.21-pre5-akpm 7 138.1 307.2 116.8
2.4.21-pre5-jam1 7 136.5 310.7 112.3
2.5.66-mjb1 7 136.4 311.1 133.1
2.4.21-pre5-ac3 7 135.6 312.9 119.4
2.2.23 7 132.3 320.6 124.8
2.5.66 7 130.8 324.3 138.1
2.5.66-ac1 7 127.6 332.6 136.8
2.5.66-mm1 7 127.2 333.4 137.9
2.4.21-pre5-akpm 8 161.2 300.7 126.7
2.4.21-pre5 8 160.9 301.4 125.1
2.4.21-pre5-ac3 8 159.4 304.1 136.8
2.4.21-pre5-jam1 8 157.4 308.0 129.1
2.5.66-mjb1 8 157.4 308.0 144.5
2.4.21-pre5aa2 8 156.2 310.4 128.3
2.2.23 8 153.4 316.1 138.0
2.5.66 8 143.3 338.3 149.7
2.5.66-ac1 8 135.6 357.5 149.7
2.5.66-mm1 8 131.8 367.9 159.8
AIM7 shared workload
This is sort of like the "university" workload.
kernel Tasks Jobs/Min Real CPU
2.4.21-pre5-jam1 8 505.6 92.1 79.8
2.4.21-pre5 8 505.1 92.2 82.2
2.4.21-pre5-ac3 8 480.2 97.0 82.0
2.4.21-pre5aa2 8 470.4 99.0 79.1
2.5.66-mm1 8 459.1 101.4 85.0
2.5.66-mjb1 8 437.1 106.5 83.0
2.5.66 8 434.8 107.1 83.7
2.5.66-ac1 8 433.1 107.5 85.3
2.4.21-pre5 16 564.5 165.0 155.0
2.4.21-pre5-jam1 16 562.8 165.5 152.9
2.4.21-pre5aa2 16 560.1 166.2 148.9
2.4.21-pre5-ac3 16 547.2 170.2 153.0
2.5.66-ac1 16 520.1 179.1 160.3
2.5.66-mjb1 16 512.9 181.6 156.3
2.5.66 16 509.1 182.9 159.1
2.5.66-mm1 16 505.3 184.3 160.6
2.4.21-pre5aa2 24 587.2 237.9 218.9
2.4.21-pre5 24 586.3 238.2 228.2
2.4.21-pre5-jam1 24 583.7 239.3 223.3
2.4.21-pre5-ac3 24 579.8 240.9 224.6
2.5.66 24 564.3 247.5 233.2
2.5.66-ac1 24 553.1 252.6 235.0
2.5.66-mjb1 24 549.0 254.4 229.3
2.5.66-mm1 24 540.3 258.5 236.2
2.4.21-pre5-jam1 32 597.5 311.7 293.1
2.4.21-pre5 32 594.7 313.2 303.1
2.4.21-pre5aa2 32 584.9 318.4 288.4
2.5.66 32 577.0 322.8 305.2
2.4.21-pre5-ac3 32 575.7 323.5 295.2
2.5.66-mjb1 32 570.9 326.2 303.8
2.5.66-mm1 32 567.8 328.0 312.2
2.5.66-ac1 32 553.9 336.2 310.3
2.4.21-pre5-ac3 40 616.0 377.9 366.4
2.4.21-pre5-jam1 40 605.4 384.6 363.6
2.4.21-pre5 40 603.5 385.8 375.7
2.4.21-pre5aa2 40 602.8 386.2 357.4
2.5.66 40 570.1 408.4 379.8
2.5.66-mjb1 40 568.5 409.5 375.6
2.5.66-mm1 40 566.3 411.1 387.8
2.5.66-ac1 40 552.3 421.5 384.0
2.4.21-pre5aa2 48 617.2 452.6 428.0
2.4.21-pre5-jam1 48 610.7 457.4 435.8
2.4.21-pre5 48 610.4 457.7 447.6
2.4.21-pre5-ac3 48 605.7 461.2 437.2
2.5.66 48 589.8 473.7 455.8
2.5.66-mjb1 48 582.0 480.0 447.5
2.5.66-mm1 48 578.3 483.1 463.0
2.5.66-ac1 48 570.0 490.1 458.9
2.4.21-pre5-ac3 56 620.4 525.3 509.5
2.4.21-pre5aa2 56 618.2 527.2 497.9
2.4.21-pre5 56 614.3 530.6 520.5
2.4.21-pre5-jam1 56 612.1 532.5 505.2
2.5.66-mjb1 56 592.0 550.6 522.6
2.5.66-mm1 56 587.8 554.4 536.6
2.5.66 56 574.9 567.0 527.5
2.5.66-ac1 56 573.1 568.7 532.6
2.4.21-pre5-ac3 64 630.2 591.1 580.4
2.4.21-pre5 64 616.6 604.1 594.0
2.4.21-pre5aa2 64 614.0 606.6 566.4
2.4.21-pre5-jam1 64 611.8 608.8 576.1
2.5.66-mjb1 64 596.0 625.0 595.3
2.5.66-mm1 64 590.2 631.1 610.5
2.5.66 64 585.3 636.4 599.9
2.5.66-ac1 64 578.0 644.4 606.1
dbench ext2 64 processes Average 5 runs
2.4.21-pre5 25.70 MB/second
2.5.66-mm1 25.46
2.4.21-pre5-ac3 24.89
2.5.66 23.74
2.4.21-pre5-akpm 23.50
2.5.66-ac1 23.26
2.5.66-mjb1 23.12
2.4.21-pre5-jam1 23.06
2.4.21-pre5aa2 21.72
2.2.23 12.81
dbench ext2 32 processes Average 5 runs
2.4.21-pre5aa2 33.87 MB/sec
2.4.21-pre5-ac3 32.69
2.4.21-pre5 31.57
2.4.21-pre5-jam1 30.70
2.4.21-pre5-akpm 30.66
2.5.66-mm1 30.54
2.5.66-mjb1 29.99
2.5.66 29.41
2.5.66-ac1 28.90
2.2.23 18.63
L M B E N C H 2 . 0 S U M M A R Y
------------------------------------
Below is the average of 25 runs. That's why things like page fault
show a decimal point where normally lmbench doesn't. BTW, would anyone
be interested in seeing lmbench for a quad xeon running Linux, and
a couple of Sun SMP boxes running Solaris?
The report comes from a tweaked lmbench getsummary script
http://home.earthlink.net/~rwhron/kernel/lmbench-getsummary.diff
and a script to summarize multiple lmbench runs:
http://home.earthlink.net/~rwhron/kernel/lmbsum
Processor, Processes - times in microseconds - smaller is better
----------------------------------------------------------------
null null open signal signal fork execve /bin/sh
kernel call I/O stat fstat close install handle process process process
--------------------- ------- ------- ------- ------- ------- ------- ------- ------- ------- -------
2.2.23 0.42 0.80455 4.76 1.24 5.77 1.43 2.74 788 2303 30829
2.4.21-pre5-ac3 0.61 0.72046 4.86 1.04 6.12 1.42 2.98 920 2573 11868
2.4.21-pre5-akpm 0.61 0.73119 3.32 1.02 5.28 1.41 3.16 865 2421 11636
2.4.21-pre5-jam1 0.61 0.72972 3.54 1.02 4.78 1.42 3.33 907 2344 10972
2.4.21-pre5 0.62 0.75284 3.18 1.02 5.35 1.41 3.25 927 2559 11884
2.4.21-pre5aa2 0.62 0.71754 3.63 1.03 5.30 1.41 3.34 861 2269 10536
2.5.66-mjb1 0.63 1.04734 3.58 1.46 5.77 1.89 3.95 1113 2727 12123
2.5.66 0.64 0.98871 4.29 1.42 6.78 1.91 3.88 1218 2898 12852
2.5.66-ac1 0.64 0.97472 3.90 1.70 6.37 1.92 3.91 1253 3213 13741
2.5.66-mm1 0.71 1.07320 3.81 1.53 7.49 2.01 3.96 1610 3850 14409
File select - times in microseconds - smaller is better
-------------------------------------------------------
select select select select select select select select
kernel 10 fd 100 fd 250 fd 500 fd 10 tcp 100 tcp 250 tcp 500 tcp
--------------------- ------- ------- ------- ------- ------- ------- ------- -------
2.2.23 3.75 12.67 28.53 55.10 6.27 35.2031 84.5070 171.545
2.4.21-pre5 3.87 20.06 46.68 91.25 5.70 38.0569 92.3951 182.674
2.4.21-pre5-ac3 3.87 22.55 47.01 91.34 5.86 41.2249 92.3680 181.776
2.4.21-pre5aa2 3.91 20.40 47.23 92.76 5.73 38.2701 94.0646 181.904
2.4.21-pre5-jam1 3.92 19.99 46.79 92.03 6.00 37.8522 91.4027 180.673
2.5.66-mjb1 3.93 18.46 44.01 82.71 5.93 35.8576 89.0096 169.001
2.4.21-pre5-akpm 4.10 20.28 46.70 91.42 5.93 39.3969 92.8172 182.095
2.5.66-mm1 4.15 21.32 49.67 84.47 5.92 40.5411 88.1585 170.195
2.5.66-ac1 4.49 18.85 43.20 84.02 6.22 35.7425 85.0179 171.786
2.5.66 5.10 18.83 43.21 83.73 6.79 35.8677 86.1015 168.798
Context switching with 64K - times in microseconds - smaller is better
----------------------------------------------------------------------
2proc/64k 4proc/64k 8proc/64k 16prc/64k 32prc/64k 64prc/64k 96prc/64k
kernel ctx swtch ctx swtch ctx swtch ctx swtch ctx swtch ctx swtch ctx swtch
-------------------------------- --------- --------- --------- --------- --------- ---------
2.2.23 177.20 191.20 202.63 223.87 248.23 264.09 267.52
2.4.21-pre5aa2 188.55 200.63 215.67 232.20 247.95 259.31 263.36
2.4.21-pre5-akpm 190.82 200.21 212.73 228.85 247.61 261.87 266.53
2.5.66-mjb1 191.06 201.22 213.78 228.63 250.67 266.76 270.99
2.4.21-pre5-ac3 191.98 199.54 209.23 223.86 244.86 259.73 263.87
2.5.66 194.97 201.79 214.86 234.38 254.90 270.24 274.87
2.4.21-pre5 195.23 202.02 212.71 226.13 248.11 263.87 268.53
2.4.21-pre5-jam1 197.55 206.01 215.67 230.23 246.80 259.25 263.56
2.5.66-ac1 203.93 208.69 215.60 233.49 253.35 269.42 273.94
2.5.66-mm1 246.03 246.87 245.57 250.67 257.31 265.59 270.49
File create/delete and VM system latencies in microseconds - smaller is better
----------------------------------------------------------------------------
0K 0K 1K 1K 4K 4K Mmap Prot Page
kernel Create Delete Create Delete Create Delete Latency Fault Fault
------------------------------ ------- ------- ------- ------- ------- ------- ------ ------
2.4.21-pre5-ac3 63.5 15.4 139.7 27.0 146.7 27.0 2630 1.34 5.2
2.4.21-pre5-akpm 64.1 14.1 139.5 24.3 144.3 24.3 2541 0.99 5.4
2.4.21-pre5 64.8 13.5 134.8 24.8 140.5 25.0 2593 0.84 5.1
2.5.66-mjb1 72.4 14.4 154.3 27.8 158.6 27.8 3621 0.66 8.8
2.5.66 78.2 15.1 161.0 27.9 166.9 27.9 3887 0.78 8.6
2.5.66-ac1 83.4 16.7 166.9 34.3 170.8 34.3 3942 1.25 15.2
2.5.66-mm1 89.3 17.0 182.3 35.4 188.0 35.4 4413 0.91 9.5
2.4.21-pre5aa2 89.7 14.2 165.3 27.3 174.3 27.3 2480 1.05 5.9
2.4.21-pre5-jam1 91.4 14.0 167.7 27.9 172.5 27.9 2508 0.84 6.0
2.2.23 141.3 21.7 207.5 27.5 215.4 27.4 64369 0.89 1246.0
*Local* Communication latencies in microseconds - smaller is better
-------------------------------------------------------------------
kernel Pipe AF/Unix UDP RPC/UDP TCP RPC/TCP TCPconn
------------------------------ ------- ------- ------- ------- ------- -------
2.4.21-pre5-ac3 10.18 21.91 39.8783 144.699 71.9270 214.945 321.89
2.4.21-pre5-akpm 10.29 18.64 44.3550 125.465 67.8534 175.328 313.69
2.2.23 10.87 33.38 48.8845 132.208 62.5111 170.780 252.82
2.4.21-pre5-jam1 11.07 24.00 45.9692 143.535 72.4726 188.593 329.04
2.4.21-pre5 11.97 21.92 40.1205 127.589 65.7712 171.820 314.04
2.4.21-pre5aa2 12.39 24.94 49.6436 135.613 77.8481 190.294 310.32
2.5.66-mjb1 12.57 27.19 66.9500 162.132 94.7787 199.190 347.71
2.5.66-ac1 14.36 30.92 83.6887 175.328 111.306 215.802 382.38
2.5.66-mm1 16.00 23.72 66.1233 172.186 88.1937 218.058 375.84
2.5.66 16.09 26.41 62.4870 145.381 91.7622 181.263 351.05
*Local* Communication bandwidths in MB/s - bigger is better
-----------------------------------------------------------
2.5.66-mjb1 has the one liner Linus posted in
http://marc.theaimsgroup.com/?l=linux-kernel&m=104849532307747&w=2
which may have helped TCP bandwidth.
Memory read bandwidth appears directly related with HZ=100. -mjb1 had HZ=100.
File Mmap Bcopy Bcopy Memory Memory
kernel Pipe AF/Unix TCP reread reread (libc) (hand) read write
------------------------------ ------- ------- ------- ------- ------- ------- ------- -------
2.2.23 115.6 38.4 39.1 62.6 237.9 61.1 61.1 237.8 88.0
2.4.21-pre5-akpm 68.4 41.7 48.5 58.9 237.7 59.8 59.8 237.5 84.7
2.4.21-pre5aa2 66.1 49.5 46.2 59.0 237.8 60.1 60.1 237.6 84.8
2.4.21-pre5-ac3 64.8 41.7 50.4 60.5 237.8 60.2 60.2 237.6 84.8
2.4.21-pre5 64.6 42.0 42.3 59.9 237.6 60.2 60.2 237.4 84.8
2.4.21-pre5-jam1 64.6 42.4 54.7 59.2 237.7 59.9 59.9 237.6 84.8
2.5.66-mjb1 64.6 44.7 46.6 59.7 237.7 60.8 60.8 237.6 86.2
2.5.66-ac1 63.2 43.5 29.1 57.2 232.8 59.0 59.0 232.7 83.8
2.5.66 63.1 42.7 31.5 57.4 233.0 59.8 59.7 232.9 85.2
2.5.66-mm1 52.0 34.3 24.9 50.3 232.6 60.1 60.1 232.3 86.6
*Local* More Communication bandwidths in MB/s - bigger is better
----------------------------------------------------------------
File Mmap Aligned Partial Partial Partial Partial
OS open open Bcopy Bcopy Mmap Mmap Mmap Bzero
close close (libc) (hand) read write rd/wrt copy HTTP
------------------------------ ------- ------- ------- ------- ------- ------- ------- -------
2.2.23 63.4 199.1 60.5 67.6 247.4 88.0 88.0 88.0 3.85
2.4.21-pre5-ac3 60.7 196.8 59.6 66.8 247.3 84.8 84.8 84.8 3.44
2.5.66-mjb1 60.6 182.3 60.2 67.3 247.3 86.2 86.2 86.2 3.19
2.4.21-pre5 60.2 196.5 59.6 66.8 247.2 84.8 84.8 84.8 3.45
2.4.21-pre5-jam1 58.9 197.2 59.4 66.6 247.3 84.8 84.8 84.8 3.53
2.4.21-pre5-akpm 58.7 189.8 59.2 66.5 247.2 84.7 84.7 84.7 3.54
2.4.21-pre5aa2 58.7 194.2 59.6 66.8 247.4 84.8 84.8 84.8 3.44
2.5.66 58.1 177.4 59.2 65.9 242.5 85.2 85.2 85.2 3.15
2.5.66-ac1 57.2 177.8 58.5 65.3 242.4 83.8 83.8 83.8 2.78
2.5.66-mm1 50.2 173.4 59.6 66.3 242.0 86.6 86.7 86.6 3.00
Memory latencies in nanoseconds - smaller is better
---------------------------------------------------
kernel Mhz L1 $ L2 $ Main mem
-------------------- ----- ------- ------- ---------
2.2.23 476 4.20 185.20 261.5
2.5.66-mjb1 476 4.20 186.90 262.1
2.4.21-pre5-akpm 476 4.20 193.94 262.1
2.4.21-pre5-jam1 476 4.20 194.20 262.0
2.5.66 476 4.25 194.47 266.8
2.4.21-pre5aa2 476 4.20 195.28 261.9
2.4.21-pre5-ac3 476 4.20 198.24 262.1
2.4.21-pre5 476 4.20 198.68 262.1
2.5.66-ac1 476 4.26 202.47 266.8
2.5.66-mm1 476 4.25 229.87 267.8
File size = 2048 MB
Block size = 4K
Read, write, and seek rates in MB/sec.
Latency in milliseconds.
Percent of requests that took longer than 2 and 10 seconds.
Anticipatory scheduler in 2.5.66-mm1 helps tiobench sequential reads.
Sequential Reads
Num Avg Maximum Lat% Lat% CPU
Kernel Thr Rate (CPU%) Latency Latency >2s >10s Eff
-------------------- --- ------------------------------------------------------------
2.4.21-pre5aa2 8 15.71 39.09% 5.896 3241.42 0.00000 0.00000 40
2.4.21-pre5-jam1 8 14.65 33.99% 6.325 3531.40 0.00000 0.00000 43
2.5.66-mm1 8 13.12 18.67% 7.014 1489.04 0.00000 0.00000 70
2.4.21-pre5-akpm 8 10.64 28.17% 8.685 868.90 0.00000 0.00000 38
2.4.21-pre5-ac3 8 10.44 28.01% 8.921 731.74 0.00000 0.00000 37
2.4.21-pre5 8 10.35 27.49% 8.830 1189.25 0.00000 0.00000 38
2.5.66 8 9.03 11.12% 10.314 678.88 0.00000 0.00000 81
2.5.66-mjb1 8 9.03 11.15% 10.310 965.88 0.00000 0.00000 81
2.5.66-ac1 8 9.01 14.70% 10.305 1178.45 0.00000 0.00000 61
2.2.23 8 7.31 50.72% 12.582 1213.59 0.00000 0.00000 14
2.4.21-pre5aa2 16 15.57 40.86% 11.840 9568.78 0.00497 0.00000 38
2.4.21-pre5-jam1 16 14.45 33.24% 12.694 7311.16 0.00420 0.00000 43
2.5.66-mm1 16 13.20 19.48% 13.913 2943.87 0.00000 0.00000 68
2.4.21-pre5-akpm 16 10.49 27.42% 17.544 1331.26 0.00000 0.00000 38
2.4.21-pre5-ac3 16 10.36 27.05% 17.940 1047.48 0.00000 0.00000 38
2.4.21-pre5 16 10.31 27.05% 17.842 1485.49 0.00000 0.00000 38
2.5.66-ac1 16 9.59 15.10% 19.386 1502.64 0.00000 0.00000 63
2.5.66 16 9.46 11.61% 19.654 1522.42 0.00000 0.00000 81
2.5.66-mjb1 16 9.24 11.08% 20.087 1521.03 0.00000 0.00000 83
2.2.23 16 7.20 49.54% 25.354 1969.50 0.00000 0.00000 15
2.4.21-pre5aa2 32 15.11 38.81% 24.106 18274.96 0.29869 0.00000 39
2.4.21-pre5-jam1 32 13.71 31.77% 26.502 16109.75 0.34523 0.00000 43
2.5.66-mm1 32 13.01 20.28% 28.125 5613.49 0.00000 0.00000 64
2.4.21-pre5-ac3 32 10.09 25.99% 28.083 367251.27 0.00687 0.00687 39
2.4.21-pre5-akpm 32 10.06 25.66% 27.736 345319.64 0.00763 0.00763 39
2.4.21-pre5 32 9.99 25.59% 28.148 351517.19 0.00820 0.00820 39
2.5.66-ac1 32 9.88 15.74% 37.601 1669.63 0.00000 0.00000 63
2.5.66 32 9.76 11.95% 38.023 1748.96 0.00000 0.00000 82
2.5.66-mjb1 32 9.66 12.11% 38.290 1735.93 0.00000 0.00000 80
2.2.23 32 7.08 48.42% 52.057 5746.36 0.00019 0.00000 15
2.4.21-pre5aa2 64 14.58 36.03% 48.962 34728.57 1.41545 0.00458 40
2.5.66-mm1 64 12.99 21.54% 55.513 13587.20 1.66855 0.00000 60
2.4.21-pre5-jam1 64 12.97 27.85% 55.227 36906.52 1.58978 0.00495 47
2.4.21-pre5-ac3 64 10.04 25.96% 47.945 487069.58 0.02747 0.02689 39
2.4.21-pre5 64 9.87 25.28% 47.652 532973.74 0.02689 0.02669 39
2.4.21-pre5-akpm 64 9.83 24.86% 47.753 529269.64 0.02728 0.02688 40
2.5.66 64 8.91 10.94% 82.612 3121.22 0.00000 0.00000 81
2.5.66-ac1 64 8.73 13.97% 85.225 3436.39 0.00000 0.00000 62
2.5.66-mjb1 64 8.16 9.99% 91.086 3654.62 0.00000 0.00000 82
2.2.23 64 5.20 20.95% 137.069 104521.20 1.02176 0.06485 25
2.4.21-pre5aa2 128 12.65 31.91% 96.873 71395.60 1.59683 0.13085 40
2.5.66-mm1 128 11.92 19.55% 107.252 136468.39 1.99680 0.15888 61
2.4.21-pre5-jam1 128 11.91 25.93% 113.047 72717.60 1.77384 0.25444 46
2.4.21-pre5-ac3 128 9.76 25.20% 87.269 564245.01 0.06294 0.06294 39
2.4.21-pre5 128 9.48 24.08% 88.698 558980.86 0.06218 0.06218 39
2.4.21-pre5-akpm 128 9.48 24.04% 89.516 571921.71 0.06256 0.06256 39
2.5.66-ac1 128 8.44 13.71% 165.441 100118.04 0.20714 0.17644 62
2.5.66 128 8.39 10.80% 167.630 93104.09 0.30442 0.19664 78
2.5.66-mjb1 128 7.55 9.80% 185.803 102350.53 0.39024 0.27237 77
2.2.23 128 5.15 37.27% 244.276 265334.68 2.03476 0.48886 14
Random Reads
Num Avg Maximum Lat% Lat% CPU
Kernel Thr Rate (CPU%) Latency Latency >2s >10s Eff
-------------------- --- ------------------------------------------------------------
2.4.21-pre5-jam1 8 0.55 1.75% 167.912 499.46 0.00000 0.00000 31
2.4.21-pre5aa2 8 0.55 1.40% 167.942 521.14 0.00000 0.00000 39
2.5.66-ac1 8 0.55 1.87% 159.920 516.81 0.00000 0.00000 30
2.4.21-pre5-ac3 8 0.54 1.39% 170.808 523.61 0.00000 0.00000 39
2.4.21-pre5-akpm 8 0.54 1.54% 170.499 529.23 0.00000 0.00000 35
2.4.21-pre5 8 0.53 1.26% 170.704 518.21 0.00000 0.00000 42
2.5.66-mm1 8 0.52 1.94% 162.919 1414.08 0.00000 0.00000 27
2.5.66 8 0.49 1.54% 180.655 658.09 0.00000 0.00000 32
2.2.23 8 0.48 0.96% 190.043 847.36 0.00000 0.00000 50
2.5.66-mjb1 8 0.46 1.17% 194.999 931.34 0.00000 0.00000 39
2.5.66-ac1 16 0.59 1.98% 297.030 1050.86 0.00000 0.00000 30
2.4.21-pre5-akpm 16 0.58 1.62% 310.519 931.69 0.00000 0.00000 36
2.4.21-pre5-jam1 16 0.58 1.68% 312.564 974.35 0.00000 0.00000 34
2.4.21-pre5aa2 16 0.58 1.89% 312.557 969.32 0.00000 0.00000 31
2.4.21-pre5 16 0.57 1.39% 314.526 992.64 0.00000 0.00000 41
2.4.21-pre5-ac3 16 0.57 1.47% 317.808 978.50 0.00000 0.00000 39
2.5.66-mm1 16 0.54 2.00% 323.277 1910.14 0.00000 0.00000 27
2.5.66 16 0.53 1.62% 331.967 1319.00 0.00000 0.00000 33
2.2.23 16 0.50 0.96% 359.150 1622.15 0.00000 0.00000 52
2.5.66-mjb1 16 0.49 1.43% 360.497 1420.27 0.00000 0.00000 34
2.4.21-pre5-jam1 32 0.61 1.97% 581.470 1736.66 0.00000 0.00000 31
2.4.21-pre5aa2 32 0.61 1.62% 584.397 1732.22 0.00000 0.00000 37
2.4.21-pre5 32 0.60 1.70% 536.530 1594.51 0.00000 0.00000 36
2.4.21-pre5-ac3 32 0.60 1.73% 541.247 1741.38 0.00000 0.00000 35
2.4.21-pre5-akpm 32 0.60 1.87% 538.854 1663.76 0.00000 0.00000 32
2.5.66-ac1 32 0.60 2.12% 588.975 1509.44 0.00000 0.00000 28
2.5.66 32 0.56 1.76% 617.477 2635.82 0.00000 0.00000 32
2.5.66-mm1 32 0.56 2.13% 627.621 3783.47 0.00000 0.00000 26
2.2.23 32 0.53 1.29% 675.727 2773.77 0.00000 0.00000 41
2.5.66-mjb1 32 0.53 2.01% 665.634 2662.49 0.00000 0.00000 26
2.4.21-pre5 64 0.63 1.80% 989.186 2733.46 0.00000 0.00000 35
2.4.21-pre5-ac3 64 0.63 2.21% 981.317 22792.03 0.05040 0.05040 29
2.4.21-pre5-akpm 64 0.63 1.83% 979.118 2672.51 0.00000 0.00000 34
2.4.21-pre5-jam1 64 0.60 1.93% 1153.455 2700.20 0.00000 0.00000 31
2.4.21-pre5aa2 64 0.60 1.87% 1150.143 2654.97 0.00000 0.00000 32
2.5.66-mm1 64 0.58 2.38% 1140.931 8080.52 0.88206 0.00000 24
2.2.23 64 0.55 1.36% 1249.718 4557.56 0.00000 0.00000 40
2.5.66-ac1 64 0.49 1.79% 1455.626 2113.49 0.00000 0.00000 27
2.5.66 64 0.48 1.56% 1490.181 4484.06 0.00000 0.00000 31
2.5.66-mjb1 64 0.41 1.50% 1717.455 4176.38 0.00000 0.00000 28
2.4.21-pre5-ac3 128 0.64 1.98% 1399.103 59500.23 4.58669 4.58669 32
2.4.21-pre5 128 0.63 2.13% 1400.224 60771.56 4.43548 4.43548 29
2.4.21-pre5-akpm 128 0.62 2.37% 1426.698 60503.21 4.63709 4.63709 26
2.4.21-pre5aa2 128 0.61 1.89% 2075.855 3971.94 0.00000 0.00000 32
2.4.21-pre5-jam1 128 0.60 2.31% 2197.111 4131.21 0.00000 0.00000 26
2.5.66-mm1 128 0.58 2.33% 2113.473 23135.97 28.37702 0.02520 25
2.2.23 128 0.57 1.74% 2185.172 10879.15 8.26613 0.00000 33
2.5.66-ac1 128 0.47 1.84% 2932.536 3708.91 0.00000 0.00000 25
2.5.66 128 0.45 1.56% 2645.401 8088.29 12.62601 0.00000 29
2.5.66-mjb1 128 0.40 1.42% 3111.081 10415.63 16.80948 0.00000 29
Sequential Writes
2.5.66-mm1 and 2.5.66-ac1 have lower max latency for tiobench sequential writes.
Num Avg Maximum Lat% Lat% CPU
Kernel Thr Rate (CPU%) Latency Latency >2s >10s Eff
-------------------- --- ------------------------------------------------------------
2.5.66 8 18.78 85.25% 3.424 178894.38 0.01526 0.00534 22
2.5.66-mjb1 8 18.59 82.74% 3.947 32026.28 0.01240 0.00382 22
2.4.21-pre5-ac3 8 18.48 83.60% 4.554 5317.48 0.00114 0.00000 22
2.5.66-mm1 8 18.43 83.78% 3.811 31290.24 0.01583 0.00477 22
2.4.21-pre5aa2 8 18.39 81.53% 4.382 5686.84 0.00172 0.00000 23
2.4.21-pre5-jam1 8 18.28 77.44% 4.332 5091.92 0.00153 0.00000 24
2.4.21-pre5-akpm 8 18.10 76.02% 4.365 3173.96 0.00000 0.00000 24
2.4.21-pre5 8 17.87 75.40% 4.364 2862.92 0.00000 0.00000 24
2.5.66-ac1 8 17.52 86.38% 3.908 124582.65 0.00972 0.00458 20
2.2.23 8 8.92 54.70% 9.598 26099.53 0.06542 0.00000 16
2.5.66-mjb1 16 18.75 83.91% 7.103 53969.82 0.03853 0.00210 22
2.5.66 16 18.68 85.47% 7.257 132124.46 0.04844 0.00210 22
2.4.21-pre5-ac3 16 18.55 84.55% 8.784 11323.07 0.29908 0.00000 22
2.5.66-mm1 16 18.39 83.80% 8.254 21452.44 0.02784 0.00000 22
2.4.21-pre5aa2 16 18.21 82.65% 8.626 8279.10 0.30594 0.00000 22
2.4.21-pre5-jam1 16 17.66 73.84% 8.395 18944.04 0.01220 0.00019 24
2.5.66-ac1 16 17.45 85.34% 8.299 26440.98 0.04025 0.00000 20
2.4.21-pre5-akpm 16 16.68 66.39% 9.409 6244.89 0.00114 0.00000 25
2.4.21-pre5 16 16.12 63.76% 9.721 6615.84 0.00324 0.00000 25
2.2.23 16 8.93 55.17% 18.526 32732.86 0.26397 0.00858 16
2.4.21-pre5-ac3 32 18.51 83.59% 16.709 24881.23 0.32043 0.00000 22
2.5.66-mjb1 32 18.46 85.50% 13.999 55092.70 0.23594 0.00439 22
2.5.66-mm1 32 18.30 83.54% 16.600 21049.50 0.36106 0.00000 22
2.5.66 32 18.22 84.32% 15.058 130920.88 0.29831 0.00382 22
2.4.21-pre5aa2 32 18.15 80.68% 16.412 16743.09 0.30898 0.00000 22
2.5.66-ac1 32 17.40 84.82% 17.537 24589.53 0.43659 0.00000 21
2.4.21-pre5-jam1 32 16.30 68.90% 17.667 27569.82 0.16537 0.00095 24
2.4.21-pre5-akpm 32 15.64 62.84% 19.920 10748.05 0.19703 0.00000 25
2.4.21-pre5 32 15.52 62.01% 19.848 12550.58 0.17472 0.00000 25
2.2.23 32 8.92 36.93% 33.105 112746.09 0.29698 0.11311 24
2.5.66-mjb1 64 18.29 86.41% 23.295 150083.73 0.21839 0.03165 21
2.4.21-pre5-ac3 64 18.28 82.52% 31.450 52962.09 0.31834 0.18844 22
2.5.66 64 18.16 85.88% 23.227 155547.87 0.26284 0.03089 21
2.5.66-mm1 64 18.16 84.07% 23.856 126201.26 0.29869 0.02651 22
2.4.21-pre5aa2 64 17.96 78.83% 30.039 33412.35 0.29983 0.21763 23
2.5.66-ac1 64 17.35 84.57% 33.920 33212.81 0.47855 0.00476 21
2.4.21-pre5-akpm 64 14.76 57.22% 40.559 19676.52 1.17893 0.00000 26
2.4.21-pre5-jam1 64 14.73 62.60% 39.467 32992.99 1.13773 0.00972 24
2.4.21-pre5 64 14.61 56.60% 40.435 20848.72 1.16977 0.00000 26
2.2.23 64 8.85 42.47% 59.342 171877.37 0.31529 0.22011 21
2.4.21-pre5aa2 128 17.78 78.47% 51.778 81396.58 0.29182 0.25692 23
2.4.21-pre5-ac3 128 17.49 78.50% 55.521 111419.23 0.36220 0.27943 22
2.5.66-ac1 128 17.33 86.10% 58.584 135422.32 0.50907 0.27046 20
2.5.66 128 16.41 78.56% 59.396 203756.11 0.60959 0.18025 21
2.5.66-mjb1 128 15.91 79.57% 64.581 173801.79 0.73185 0.16843 20
2.5.66-mm1 128 15.83 70.45% 62.154 196987.22 0.67730 0.17338 22
2.4.21-pre5-akpm 128 14.13 54.60% 80.088 37666.28 1.42287 0.00725 26
2.4.21-pre5 128 14.09 55.01% 79.247 39973.07 1.40095 0.00667 26
2.4.21-pre5-jam1 128 13.41 59.71% 84.153 45593.95 1.38131 0.02937 22
2.2.23 128 8.77 59.92% 90.360 243194.33 0.36029 0.26818 15
Random Writes
Num Avg Maximum Lat% Lat% CPU
Kernel Thr Rate (CPU%) Latency Latency >2s >10s Eff
-------------------- --- ------------------------------------------------------------
2.2.23 8 0.73 5.50% 7.244 5326.51 0.00000 0.00000 13
2.4.21-pre5aa2 8 0.72 3.90% 0.281 290.40 0.00000 0.00000 19
2.4.21-pre5-jam1 8 0.67 3.94% 0.278 291.25 0.00000 0.00000 17
2.5.66-ac1 8 0.64 2.04% 0.553 599.56 0.00000 0.00000 31
2.4.21-pre5-akpm 8 0.62 1.16% 0.984 518.14 0.00000 0.00000 53
2.4.21-pre5 8 0.61 1.11% 1.304 246.92 0.00000 0.00000 55
2.5.66-mm1 8 0.58 1.79% 16.630 1703.67 0.00000 0.00000 32
2.4.21-pre5-ac3 8 0.56 1.20% 0.207 1.88 0.00000 0.00000 47
2.5.66-mjb1 8 0.55 1.76% 18.296 1936.69 0.00000 0.00000 31
2.5.66 8 0.52 1.60% 19.003 1619.82 0.00000 0.00000 33
2.4.21-pre5aa2 16 0.75 4.30% 0.317 401.99 0.00000 0.00000 17
2.2.23 16 0.73 6.05% 15.480 5130.98 0.00000 0.00000 12
2.4.21-pre5-jam1 16 0.68 3.63% 0.197 4.22 0.00000 0.00000 19
2.5.66-ac1 16 0.66 2.17% 0.740 1427.84 0.00000 0.00000 30
2.4.21-pre5 16 0.62 1.19% 1.508 409.31 0.00000 0.00000 52
2.4.21-pre5-akpm 16 0.62 1.32% 0.600 535.15 0.00000 0.00000 47
2.4.21-pre5-ac3 16 0.58 1.19% 0.208 2.45 0.00000 0.00000 48
2.5.66-mm1 16 0.58 1.76% 33.893 3109.52 0.00000 0.00000 33
2.5.66-mjb1 16 0.54 1.67% 36.083 2115.71 0.00000 0.00000 32
2.5.66 16 0.53 1.64% 35.789 4007.02 0.00000 0.00000 32
2.4.21-pre5aa2 32 0.79 4.85% 0.210 4.23 0.00000 0.00000 16
2.4.21-pre5-jam1 32 0.69 4.16% 0.287 340.14 0.00000 0.00000 17
2.2.23 32 0.68 8.55% 125.365 5932.63 0.02500 0.00000 8
2.5.66-ac1 32 0.66 2.13% 1.317 2007.86 0.00000 0.00000 31
2.4.21-pre5-akpm 32 0.65 1.36% 0.715 683.88 0.00000 0.00000 48
2.4.21-pre5 32 0.63 1.29% 1.745 584.37 0.00000 0.00000 49
2.4.21-pre5-ac3 32 0.62 1.31% 0.206 1.99 0.00000 0.00000 47
2.5.66-mm1 32 0.57 1.74% 71.868 5994.63 0.05000 0.00000 33
2.5.66-mjb1 32 0.55 1.69% 68.461 2848.62 0.00000 0.00000 32
2.5.66 32 0.53 1.64% 70.277 4485.75 0.00000 0.00000 32
2.4.21-pre5aa2 64 0.80 4.66% 0.206 3.93 0.00000 0.00000 17
2.4.21-pre5-jam1 64 0.72 4.26% 0.195 4.85 0.00000 0.00000 17
2.4.21-pre5-ac3 64 0.67 1.54% 0.208 1.90 0.00000 0.00000 43
2.4.21-pre5-akpm 64 0.67 1.31% 0.370 425.43 0.00000 0.00000 51
2.2.23 64 0.66 15.71% 210.913 7787.02 0.60483 0.00000 4
2.5.66-ac1 64 0.66 2.18% 5.810 8032.91 0.15121 0.00000 30
2.4.21-pre5 64 0.65 1.35% 1.953 667.44 0.00000 0.00000 48
2.5.66-mm1 64 0.58 1.78% 128.446 8412.81 0.85686 0.00000 32
2.5.66-mjb1 64 0.55 1.95% 164.523 9248.35 0.27722 0.00000 28
2.5.66 64 0.54 1.70% 127.350 7290.77 0.07560 0.00000 32
2.4.21-pre5aa2 128 0.80 4.57% 0.205 4.40 0.00000 0.00000 18
2.4.21-pre5-ac3 128 0.72 1.73% 0.206 2.07 0.00000 0.00000 42
2.4.21-pre5-akpm 128 0.71 1.47% 0.476 679.34 0.00000 0.00000 48
2.4.21-pre5-jam1 128 0.71 4.02% 0.192 4.66 0.00000 0.00000 18
2.4.21-pre5 128 0.69 1.62% 1.249 700.81 0.00000 0.00000 43
2.5.66-ac1 128 0.69 2.34% 16.169 14838.16 0.32762 0.00000 30
2.2.23 128 0.59 46.26% 438.313 22960.69 4.68750 0.00000 1
2.5.66-mm1 128 0.57 1.90% 257.503 16533.61 3.35182 0.00000 30
2.5.66 128 0.54 1.78% 280.961 18245.53 2.19254 0.00000 30
2.5.66-mjb1 128 0.53 1.61% 317.130 17797.19 2.52016 0.00000 33
bonnie++-1.02c 9/14/2002 with a 1024 MB file.
2.5.x has lower cpu utilization for sequential block reads.
---------------------Sequential Output--------------------
-----Per Char----- ------Block------- -----Rewrite------
Kernel MB/sec %CPU Eff MB/sec %CPU Eff MB/sec %CPU Eff
2.4.21-pre5aa2 3.80 98.0 3.88 22.46 85.7 26.21 9.86 52.0 18.96
2.4.21-pre5-jam1 3.77 98.0 3.85 21.99 87.0 25.28 9.81 53.3 18.40
2.4.21-pre5 3.76 98.0 3.84 21.43 87.0 24.63 9.62 44.7 21.55
2.4.21-pre5-akpm 3.76 98.0 3.84 21.31 87.7 24.31 8.88 34.7 25.61
2.4.21-pre5-ac3 3.76 98.0 3.84 21.21 86.7 24.47 9.54 44.3 21.52
2.5.66-mjb1 3.69 97.0 3.81 20.96 87.0 24.09 7.97 32.3 24.64
2.5.66 3.66 97.0 3.78 20.55 85.0 24.18 7.37 31.7 23.27
2.5.66-mm1 3.68 97.0 3.80 20.48 85.0 24.10 7.85 28.0 28.05
2.5.66-ac1 3.63 97.0 3.74 20.43 83.7 24.42 7.80 33.3 23.39
2.2.23 2.96 73.7 4.02 9.85 59.3 16.60 4.47 87.3 5.11
-----------Sequential Input----------- ------Random-----
-----Per Char----- ------Block------- ------Seeks------
Kernel MB/sec %CPU Eff MB/sec %CPU Eff /sec %CPU Eff
2.4.21-pre5aa2 3.92 95.7 4.09 21.26 77.3 27.49 140 1.7 8402
2.4.21-pre5-jam1 3.92 96.3 4.07 21.93 82.0 26.74 136 1.7 8150
2.4.21-pre5 4.01 97.0 4.13 18.30 66.7 27.45 144 1.7 8656
2.4.21-pre5-akpm 3.86 94.0 4.11 17.19 52.3 32.85 141 2.0 7038
2.4.21-pre5-ac3 4.01 98.0 4.09 18.40 66.3 27.74 138 1.7 8270
2.5.66-mjb1 4.00 99.0 4.04 15.00 16.3 91.84 126 3.0 4210
2.5.66 3.94 98.3 4.01 14.24 17.0 83.76 137 3.0 4574
2.5.66-mm1 3.92 98.0 4.00 14.58 16.7 87.46 154 3.0 5140
2.5.66-ac1 3.98 99.0 4.02 14.64 16.3 89.65 134 4.0 3344
2.2.23 3.05 96.7 3.15 9.62 77.3 12.43 133 1.0 13329
These are all seconds to build a package, or in the case of bonnie, time to
run the benchmark. NFS numbers are missing for 2.5.66*. I believe 2.5.x
needs NFS v3.
kernel autoconf-build bonnie++-3x bonnie++nfs-3x kernel-build perl-build
2.4.21-pre5-jam1 3326(3x) 2629 4133 1557 1271 seconds - lower is better.
2.4.21-pre5-ac3 3463(3x) 2658 4181 1604 1286
2.4.21-pre5 3468(3x) 2651 na 1617 1288
2.5.66-mjb1 3675(3x) 2822 na 1558 1253
2.5.66 3772(3x) 2871 na 1636 1249
2.2.23 3777(3x) 4077 na 1593 1396
2.5.66-ac1 3846(3x) 2836 na 1680 1322
2.5.66-mm1 3947(3x) 2822 na 1621 1316
The tcp debug tweak in -mjb1 didn't seem to make a difference
for tbench.
tbench 32 processes Average 5 runs
2.4.21-pre5-jam1 18.27 MB/second
2.4.21-pre5-ac3 17.48
2.4.21-pre5aa2 16.87
2.5.66 16.18
2.5.66-mjb1 16.18
2.5.66-mm1 15.86
2.5.66-ac1 14.51
2.4.21-pre5-akpm 13.76
2.4.21-pre5 13.47
2.2.23 13.20
--
Randy Hron
http://home.earthlink.net/~rwhron/kernel/bigbox.html
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [benchmarks] on recent kernels
2003-04-01 23:29 [benchmarks] on recent kernels rwhron
@ 2003-04-03 14:07 ` Andrea Arcangeli
0 siblings, 0 replies; 2+ messages in thread
From: Andrea Arcangeli @ 2003-04-03 14:07 UTC (permalink / raw)
To: rwhron; +Cc: linux-kernel
Hi Randy,
Thank you so much again for those so helpful benchmark results.
just a few comments.
On Tue, Apr 01, 2003 at 06:29:50PM -0500, rwhron@earthlink.net wrote:
> File create/delete and VM system latencies in microseconds - smaller is better
> ----------------------------------------------------------------------------
> 0K 0K 1K 1K 4K 4K Mmap Prot Page
> kernel Create Delete Create Delete Create Delete Latency Fault Fault
> ------------------------------ ------- ------- ------- ------- ------- ------- ------ ------
> 2.4.21-pre5-ac3 63.5 15.4 139.7 27.0 146.7 27.0 2630 1.34 5.2
> 2.4.21-pre5-akpm 64.1 14.1 139.5 24.3 144.3 24.3 2541 0.99 5.4
> 2.4.21-pre5 64.8 13.5 134.8 24.8 140.5 25.0 2593 0.84 5.1
> 2.5.66-mjb1 72.4 14.4 154.3 27.8 158.6 27.8 3621 0.66 8.8
> 2.5.66 78.2 15.1 161.0 27.9 166.9 27.9 3887 0.78 8.6
> 2.5.66-ac1 83.4 16.7 166.9 34.3 170.8 34.3 3942 1.25 15.2
> 2.5.66-mm1 89.3 17.0 182.3 35.4 188.0 35.4 4413 0.91 9.5
> 2.4.21-pre5aa2 89.7 14.2 165.3 27.3 174.3 27.3 2480 1.05 5.9
> 2.4.21-pre5-jam1 91.4 14.0 167.7 27.9 172.5 27.9 2508 0.84 6.0
> 2.2.23 141.3 21.7 207.5 27.5 215.4 27.4 64369 0.89 1246.0
the reason my tree is slower in create is intentional: I drop the
negative dentries after unlink to better preserve the working set, and
to release IMHO worthless cache in smart way. I don't think it's common
to unlink and open immediatly back. the other kernels do better here
because they optimize for unlike + open. Of course also in my tree the
first open failure will trigger the reallocation of the negative dentry.
> 2.5.x has lower cpu utilization for sequential block reads.
>
> ---------------------Sequential Output--------------------
> -----Per Char----- ------Block------- -----Rewrite------
> Kernel MB/sec %CPU Eff MB/sec %CPU Eff MB/sec %CPU Eff
> 2.4.21-pre5aa2 3.80 98.0 3.88 22.46 85.7 26.21 9.86 52.0 18.96
> 2.4.21-pre5-jam1 3.77 98.0 3.85 21.99 87.0 25.28 9.81 53.3 18.40
> 2.4.21-pre5 3.76 98.0 3.84 21.43 87.0 24.63 9.62 44.7 21.55
> 2.4.21-pre5-akpm 3.76 98.0 3.84 21.31 87.7 24.31 8.88 34.7 25.61
> 2.4.21-pre5-ac3 3.76 98.0 3.84 21.21 86.7 24.47 9.54 44.3 21.52
> 2.5.66-mjb1 3.69 97.0 3.81 20.96 87.0 24.09 7.97 32.3 24.64
> 2.5.66 3.66 97.0 3.78 20.55 85.0 24.18 7.37 31.7 23.27
> 2.5.66-mm1 3.68 97.0 3.80 20.48 85.0 24.10 7.85 28.0 28.05
> 2.5.66-ac1 3.63 97.0 3.74 20.43 83.7 24.42 7.80 33.3 23.39
> 2.2.23 2.96 73.7 4.02 9.85 59.3 16.60 4.47 87.3 5.11
>
> -----------Sequential Input----------- ------Random-----
> -----Per Char----- ------Block------- ------Seeks------
> Kernel MB/sec %CPU Eff MB/sec %CPU Eff /sec %CPU Eff
> 2.4.21-pre5aa2 3.92 95.7 4.09 21.26 77.3 27.49 140 1.7 8402
> 2.4.21-pre5-jam1 3.92 96.3 4.07 21.93 82.0 26.74 136 1.7 8150
> 2.4.21-pre5 4.01 97.0 4.13 18.30 66.7 27.45 144 1.7 8656
> 2.4.21-pre5-akpm 3.86 94.0 4.11 17.19 52.3 32.85 141 2.0 7038
> 2.4.21-pre5-ac3 4.01 98.0 4.09 18.40 66.3 27.74 138 1.7 8270
> 2.5.66-mjb1 4.00 99.0 4.04 15.00 16.3 91.84 126 3.0 4210
> 2.5.66 3.94 98.3 4.01 14.24 17.0 83.76 137 3.0 4574
> 2.5.66-mm1 3.92 98.0 4.00 14.58 16.7 87.46 154 3.0 5140
> 2.5.66-ac1 3.98 99.0 4.02 14.64 16.3 89.65 134 4.0 3344
> 2.2.23 3.05 96.7 3.15 9.62 77.3 12.43 133 1.0 13329
the improvement for read contigous of my tree versus the others is
nothing here, scsi really shows the difference between my tree and
all others including 2.5. IDE is capable of 64k dma only, so the
difference is not huge in the above results. On scsi a plain bonnie
approches a 100% improvement on some high end hardware as you also can
see in bigbox.html.
thanks,
Andrea
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2003-04-03 13:56 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2003-04-01 23:29 [benchmarks] on recent kernels rwhron
2003-04-03 14:07 ` Andrea Arcangeli
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox