lkml.org 
[lkml]   [2003]   [Oct]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [BENCHMARK] I/O regression after 2.6.0-test5
From
On 28 Oct 2003 10:24:36 +1100,  Nick Piggin wrote:
> There is an #if 0 part in the patch. Could you try setting it to #if 1
> please.

Changing the #if 0 to #if 1 appears to affect AIM7 fserver the most.

Patch tested is very similar to:
http://marc.theaimsgroup.com/?l=linux-kernel&m=106729967301172&w=2

2.6.0-test9-as as-fix patch
2.6.0-test9-as1 as-fix patch with #if 1

Either way, that version of as-fix patch looks good for both
AIM7 and tiobench.

AIM7 dbase workload
kernel Tasks Jobs/Min Real CPU
2.6.0-test9-as 32 604.1 314.6 133.7
2.6.0-test9-as1 32 599.6 317.0 142.2
2.6.0-test5 32 590.2 322.0 131.9
2.6.0-test9 32 316.2 601.2 128.5

2.6.0-test9-as 64 783.3 485.3 256.3
2.6.0-test9-as1 64 776.8 489.4 257.6
2.6.0-test5 64 774.3 490.9 256.6
2.6.0-test9 64 460.9 824.8 246.2

2.6.0-test9-as 96 868.5 656.6 386.4
2.6.0-test9-as1 96 864.1 660.0 391.0
2.6.0-test5 96 859.0 663.9 389.5
2.6.0-test9 96 507.5 1123.5 368.1

2.6.0-test9-as 128 914.1 831.8 510.7
2.6.0-test9-as1 128 910.7 834.8 521.1
2.6.0-test5 128 910.3 835.2 528.6
2.6.0-test9 128 536.6 1416.9 487.0

2.6.0-test9-as 160 952.4 998.0 635.8
2.6.0-test9-as1 160 949.3 1001.2 652.2
2.6.0-test5 160 741.5 1281.7 644.8
2.6.0-test9 160 522.4 1819.2 619.9

2.6.0-test9-as1 192 981.4 1162.1 779.1
2.6.0-test9-as 192 976.6 1167.8 774.2
2.6.0-test5 192 840.0 1357.8 776.0
2.6.0-test9 192 513.5 2220.8 739.3

2.6.0-test9-as 224 1000.6 1329.7 917.0
2.6.0-test9-as1 224 994.9 1337.3 904.8
2.6.0-test5 224 986.9 1348.2 911.8
2.6.0-test9 224 503.9 2640.3 853.4

2.6.0-test9-as 256 1010.6 1504.7 1036.3
2.6.0-test9-as1 256 1009.0 1507.1 1030.0
2.6.0-test5 256 999.5 1521.3 1029.3
2.6.0-test9 256 530.8 2864.8 975.5

At low number of tasks, #if 0 does more work. As the
number of tasks goes up, having #if 1 does more work.


AIM7 fserver workload
kernel Tasks Jobs/Min Real CPU
2.6.0-test9-as 4 114.6 211.6 38.2
2.6.0-test5 4 79.7 304.0 38.3
2.6.0-test9-as1 4 79.3 305.8 38.4
2.6.0-test9 4 43.0 564.4 43.6

2.6.0-test9-as 8 175.0 277.0 65.9
2.6.0-test5 8 129.8 373.5 66.3
2.6.0-test9-as1 8 128.8 376.5 65.0
2.6.0-test9 8 61.4 789.1 68.5

2.6.0-test9-as 12 208.4 349.0 93.6
2.6.0-test5 12 164.1 443.0 93.1
2.6.0-test9-as1 12 158.4 459.1 91.6
2.6.0-test9 12 70.4 1033.1 97.6

2.6.0-test9-as 16 190.4 509.3 111.8
2.6.0-test9-as1 16 184.7 525.1 114.2
2.6.0-test5 16 173.4 559.0 116.5
2.6.0-test9 16 100.7 962.8 135.5

2.6.0-test9-as1 20 213.3 568.2 143.3
2.6.0-test9-as 20 190.3 636.8 142.8
2.6.0-test5 20 180.5 671.5 143.0
2.6.0-test9 20 113.0 1072.6 160.0

2.6.0-test9-as1 24 219.2 663.5 169.8
2.6.0-test9-as 24 193.6 751.2 167.3
2.6.0-test5 24 189.1 769.0 176.0
2.6.0-test9 24 116.8 1245.6 163.3

2.6.0-test9-as1 28 223.7 758.6 195.5
2.6.0-test9-as 28 205.7 825.0 200.6
2.6.0-test5 28 204.6 829.2 197.5
2.6.0-test9 28 132.7 1278.5 190.2

2.6.0-test9-as1 32 228.1 850.0 224.8
2.6.0-test5 32 213.7 907.4 228.9
2.6.0-test9-as 32 209.3 926.6 225.9
2.6.0-test9 32 130.4 1486.6 209.1

The regression seen after test5 is gone with the most
recent version of as-fix.

AIM7 shared workload
kernel Tasks Jobs/Min Real CPU
2.6.0-test9-as1 64 1952.3 190.8 169.6
2.6.0-test9-as 64 1936.9 192.3 170.2
2.6.0-test5 64 1888.8 197.2 168.3
2.6.0-test9 64 1050.5 354.6 165.0

2.6.0-test9-as 128 2154.9 345.7 337.2
2.6.0-test9-as1 128 2146.3 347.1 338.1
2.6.0-test5 128 2094.9 355.6 338.1
2.6.0-test9 128 1316.6 565.8 330.6

2.6.0-test9-as 192 2267.5 492.8 509.8
2.6.0-test9-as1 192 2260.9 494.2 508.6
2.6.0-test5 192 2256.0 495.3 507.8
2.6.0-test9 192 1399.0 798.7 502.9

2.6.0-test9-as1 256 2417.6 616.3 678.9
2.6.0-test5 256 2389.0 623.7 683.8
2.6.0-test9-as 256 2389.0 623.7 679.8
2.6.0-test9 256 1472.1 1012.1 672.8

2.6.0-test5 320 2457.9 757.7 856.6
2.6.0-test9-as 320 2440.5 763.1 851.0
2.6.0-test9-as1 320 2420.7 769.4 855.9
2.6.0-test9 320 1590.3 1171.1 841.0

2.6.0-test9-as1 384 2498.2 894.6 1030.3
2.6.0-test9-as 384 2487.2 898.6 1038.3
2.6.0-test5 384 2480.4 901.0 1036.5
2.6.0-test9 384 1626.5 1374.1 1022.9

2.6.0-test5 448 2529.8 1030.7 1209.2
2.6.0-test9-as 448 2514.9 1036.8 1213.3
2.6.0-test9-as1 448 2499.6 1043.1 1206.1
2.6.0-test9 448 1636.7 1593.0 1199.5

2.6.0-test9-as 512 2598.7 1146.7 1390.1
2.6.0-test9-as1 512 2581.0 1154.5 1382.3
2.6.0-test5 512 2566.2 1161.2 1386.2
2.6.0-test9 512 1650.9 1805.0 1377.2


Tiobench below is not too exciting. The main thing is the regression
seen in test[678] is gone.


tiobench-0.3.3
Unit information
================
File size = 8192 megabytes
Blk Size = 4096 bytes
Rate = megabytes per second
CPU% = percentage of CPU used during the test
Latency = milliseconds
Lat% = percent of requests that took longer than X seconds
CPU Eff = Rate divided by CPU% - throughput per cpu load
Sequential Reads ext2
Num Avg Maximum Lat%
Kernel Thr Rate (CPU%) Latency Latency >2s
---------------------- --- --------------------------------------------
2.6.0-test9-as 1 28.83 13.66% 0.404 753.63 0.00000
2.6.0-test5 1 28.55 13.89% 0.408 754.59 0.00000
2.6.0-test9-as1 1 28.47 13.76% 0.409 885.44 0.00000
2.6.0-test5 8 34.01 17.09% 2.620 834.85 0.00000
2.6.0-test9-as1 8 33.90 16.82% 2.650 943.85 0.00000
2.6.0-test9-as 8 30.59 15.02% 2.963 843.18 0.00000
2.6.0-test9-as1 16 31.26 15.30% 5.703 1078.85 0.00000
2.6.0-test5 16 30.13 15.04% 5.980 1337.69 0.00000
2.6.0-test9-as 16 26.53 13.00% 6.758 1036.38 0.00000
2.6.0-test9-as1 32 29.63 14.81% 12.057 1459.39 0.00000
2.6.0-test5 32 29.46 14.62% 11.954 1560.02 0.00000
2.6.0-test9-as 32 24.84 12.43% 14.389 1842.53 0.00000
2.6.0-test5 64 25.24 13.57% 26.901 1743.05 0.00000
2.6.0-test9-as1 64 22.77 12.34% 29.750 1783.29 0.00000
2.6.0-test9-as 64 21.53 11.64% 31.063 1901.78 0.00000
2.6.0-test5 128 10.41 9.76% 129.662 1820.78 0.00000
2.6.0-test9-as1 128 10.27 9.41% 130.097 2172.39 0.00000
2.6.0-test9-as 128 10.05 9.13% 133.312 1753.29 0.00000
2.6.0-test5 256 9.13 13.30% 305.097 1446.43 0.00000
2.6.0-test9-as 256 8.62 12.50% 322.473 1698.87 0.00000
2.6.0-test9-as1 256 8.57 12.63% 325.050 1653.00 0.00000
Random Reads ext2
Num Avg Maximum Lat%
Kernel Thr Rate (CPU%) Latency Latency >2s
---------------------- --- --------------------------------------------
2.6.0-test9-as1 1 0.98 0.99% 12.001 109.76 0.00000
2.6.0-test5 1 0.95 0.94% 12.373 114.89 0.00000
2.6.0-test9-as 1 0.81 0.80% 14.484 117.63 0.00000
2.6.0-test5 8 4.11 4.58% 19.433 137.74 0.00000
2.6.0-test9-as1 8 4.03 4.39% 20.775 177.47 0.00000
2.6.0-test9-as 8 3.61 3.79% 23.024 161.12 0.00000
2.6.0-test9-as1 16 4.58 3.69% 36.975 396.44 0.00000
2.6.0-test5 16 4.05 3.90% 42.424 236.49 0.00000
2.6.0-test9-as 16 3.33 3.02% 51.015 293.01 0.00000
2.6.0-test5 32 4.41 3.58% 70.910 516.28 0.00000
2.6.0-test9-as1 32 4.21 3.95% 80.533 451.20 0.00000
2.6.0-test9-as 32 3.26 3.13% 102.403 697.51 0.00000
2.6.0-test5 64 4.47 4.00% 137.277 1077.53 0.00000
2.6.0-test9-as 64 4.40 3.88% 134.358 1229.07 0.00000
2.6.0-test9-as1 64 4.40 3.74% 138.895 1233.53 0.00000
2.6.0-test5 128 4.86 4.83% 243.704 2287.09 0.00000
2.6.0-test9-as1 128 4.68 4.47% 243.701 2172.91 0.00000
2.6.0-test9-as 128 4.51 4.51% 249.319 2547.23 0.00000
2.6.0-test5 256 5.51 7.00% 427.503 2511.79 0.00000
2.6.0-test9-as 256 5.23 6.79% 450.963 3573.61 0.00000
2.6.0-test9-as1 256 5.01 6.61% 467.361 3305.41 0.00000
Sequential Writes ext2
Num Avg Maximum Lat%
Kernel Thr Rate (CPU%) Latency Latency >2s
---------------------- --- --------------------------------------------
2.6.0-test5 1 56.24 41.13% 0.170 3207.90 0.00000
2.6.0-test9-as1 1 55.71 39.50% 0.172 2225.67 0.00000
2.6.0-test9-as 1 55.49 39.25% 0.173 1806.75 0.00000
2.6.0-test5 8 31.86 31.70% 2.356 15355.31 0.05655
2.6.0-test9-as1 8 31.66 30.95% 2.364 14477.90 0.06008
2.6.0-test9-as 8 31.10 30.23% 2.453 14880.66 0.06271

2.6.0-test9-as1 16 32.63 31.80% 4.445 28724.28 0.08045
2.6.0-test9-as 16 32.10 31.67% 4.459 41263.17 0.07834
2.6.0-test5 16 32.07 32.22% 4.371 29214.57 0.07829

2.6.0-test9-as1 32 32.79 32.88% 8.351 70993.57 0.10638
2.6.0-test5 32 30.73 34.74% 8.843 59335.31 0.10657
2.6.0-test9-as 32 30.03 33.06% 8.852 61766.24 0.10576

2.6.0-test5 64 30.27 36.46% 16.798 119951.55 0.14372
2.6.0-test9-as1 64 28.70 34.79% 17.693 114731.30 0.15449
2.6.0-test9-as 64 27.10 32.73% 18.207 120362.26 0.15745

2.6.0-test5 128 29.98 38.86% 32.259 230569.79 0.20509
2.6.0-test9-as1 128 29.84 37.86% 31.890 224444.62 0.20604
2.6.0-test9-as 128 29.25 37.18% 32.428 210201.96 0.21028
2.6.0-test5 256 33.49 48.84% 53.571 391954.31 0.24166
2.6.0-test9-as 256 32.18 44.97% 56.356 453839.45 0.24067
2.6.0-test9-as1 256 31.23 43.69% 57.305 431916.55 0.23952

Random Writes ext2
Num Avg Maximum Lat%
Kernel Thr Rate (CPU%) Latency Latency >2s
---------------------- --- --------------------------------------------
2.6.0-test9-as 1 3.08 2.79% 0.582 25.12 0.00000
2.6.0-test5 1 2.86 2.88% 1.117 50.58 0.00000
2.6.0-test9-as1 1 2.63 2.52% 1.481 380.77 0.00000
2.6.0-test5 8 3.85 5.10% 4.885 861.30 0.00000
2.6.0-test9-as1 8 3.69 3.69% 7.495 1084.38 0.00000
2.6.0-test9-as 8 3.34 3.49% 11.409 1369.42 0.00000
2.6.0-test5 16 3.70 3.93% 17.617 1419.36 0.00000
2.6.0-test9-as1 16 3.59 3.90% 18.985 1561.92 0.00000
2.6.0-test9-as 16 3.43 3.33% 16.893 1458.50 0.00000
2.6.0-test5 32 3.72 5.11% 14.666 1191.07 0.00000
2.6.0-test9-as1 32 3.67 4.34% 13.096 1346.98 0.00000
2.6.0-test9-as 32 3.40 4.22% 16.468 1422.35 0.00000
2.6.0-test5 64 3.69 5.68% 14.445 1116.77 0.00000
2.6.0-test9-as1 64 3.53 5.15% 18.633 1450.11 0.00000
2.6.0-test9-as 64 3.39 4.76% 19.373 1659.40 0.00000
2.6.0-test5 128 3.72 6.48% 14.037 1354.59 0.00000
2.6.0-test9-as 128 3.44 5.97% 12.805 1776.43 0.00000
2.6.0-test9-as1 128 3.41 5.81% 15.659 1974.75 0.00000
2.6.0-test5 256 3.57 6.59% 13.418 1912.18 0.00000
2.6.0-test9-as1 256 3.48 6.76% 5.699 1468.80 0.00000
2.6.0-test9-as 256 3.43 6.53% 8.807 1795.06 0.00000
--
Randy Hron
http://home.earthlink.net/~rwhron/kernel/bigbox.html

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:58    [from the cache]
©2003-2011 Jasper Spaans