IOPerf results

Results were obtained using a 8 kByte buffer, 50000 IO 1 pass. The test order in that table is :

IO test on individual partition (one at a time)
One IO thread per volume, 7 volumes.
Parralele IO (10 threads per volume). One result-average shown.

Block IO Character Random

fwrite fread putc getc seek

Machine/FS KB KB/sec Real KB/sec CPU %CPU KB/sec Real KB/sec CPU %CPU KB/sec Real KB/sec CPU %CPU KB/sec Real KB/sec CPU %CPU KB/sec Real KB/sec CPU %CPU

_star_test0_ 400000 15942.61 15974.44 99.80 47169.81 47169.81 100.00 1157.24 1159.22 99.83 3344.20 3345.04 99.97 519.45 519.45 100.00

_star_test1_ 400000 15835.31 15879.32 99.72 48309.18 48309.18 100.00 1157.98 1160.09 99.82 3344.76 3346.44 99.95 519.45 519.45 100.00

_star_test2_ 400000 16077.17 16109.54 99.80 46296.30 46296.30 100.00 1155.00 1155.27 99.98 3333.61 3334.72 99.97 519.45 519.45 100.00

_star_test3_ 400000 16135.54 16214.03 99.52 46349.94 46349.94 100.00 1148.90 1150.19 99.89 3338.90 3339.73 99.97 519.45 519.45 100.00

_star_test4_ 400000 16064.26 16116.04 99.68 46082.95 46082.95 100.00 1150.32 1150.48 99.99 3330.84 3331.67 99.98 513.98 513.98 100.00

_star_test5_ 400000 16057.81 16109.54 99.68 47732.70 47732.70 100.00 1154.27 1154.50 99.98 3342.80 3343.64 99.97 515.34 515.34 100.00

_star_test6_ 400000 15847.86 16122.53 98.30 47169.81 47169.81 100.00 1140.02 1141.39 99.88 3342.53 3343.36 99.97 516.70 516.70 100.00

P_star_test0_test_ 400000 13976.24 15048.91 92.87 13093.29 29090.91 45.01 602.44 1086.28 55.46 1846.47 3195.14 57.79 279.02 489.51 57.00

_star_test0_test_ 400000 4122.22 14914.88 27.64 5622.00 25004.21 22.50 428.20 1092.32 39.20 1285.70 3204.70 40.12 354.13 495.90 71.34

Only 40000 IO (factor a 10 down, IOZone ~ 15 in file size), note the improved read-rate corresponding to the 'dome-shape' profile effect.
IO test on individual partition (one at a time)
One IO thread per volume, 4 volumes. New notation 1PV .
Parralele IO (10 threads per volume). New notation NPV .

		Block IO						Character						Random
		fwrite			fread			putc			getc			seek
Machine/FS	KB	KB/sec Real	KB/sec CPU	%CPU	KB/sec Real	KB/sec CPU	%CPU	KB/sec Real	KB/sec CPU	%CPU	KB/sec Real	KB/sec CPU	%CPU	KB/sec Real	KB/sec CPU	%CPU
_star_test0_	40000	16877.64	16877.64	100.00	83333.33	83333.33	100.00	1177.51	1178.55	99.91	3445.31	3445.31	100.00	527.87	527.87	100.00
_star_test1_	40000	15748.03	15873.02	99.21	83333.33	83333.33	100.00	1131.54	1134.75	99.72	3451.25	3451.25	100.00	527.87	527.87	100.00
_star_test2_	40000	15503.88	16064.26	96.51	85106.38	85106.38	100.00	1160.77	1160.77	100.00	3454.23	3454.23	100.00	513.98	513.98	100.00
_star_test3_	40000	16260.16	16528.93	98.37	83333.33	83333.33	100.00	1177.16	1180.64	99.71	3445.31	3445.31	100.00	500.80	500.80	100.00
1PV_star_test0_test_	40000	16194.33	16194.33	100.00	83333.33	83333.33	100.00	1093.19	1106.19	98.82	3369.84	3401.36	99.07	513.98	513.98	100.00
1PV_star_test1_test_	40000	15936.25	16000.00	99.60	78431.37	78431.37	100.00	1078.17	1094.39	98.52	3350.08	3369.84	99.41	513.98	513.98	100.00
1PV_star_test2_test_	40000	15748.03	15810.28	99.61	68965.52	68965.52	100.00	1085.48	1112.04	97.61	3347.28	3347.28	100.00	527.87	527.87	100.00
1PV_star_test3_test_	40000	13840.83	15686.27	88.24	70175.44	71428.57	98.25	1106.81	1121.08	98.73	3372.68	3372.68	100.00	513.98	513.98	100.00
NPV_star_test0_test_	40000	4813.46	14724.81	32.67	19679.77	33358.14	59.68	428.41	1098.82	38.99	1409.61	3290.86	42.84	379.89	506.07	75.17
NPV_star_test1_test_	40000	4753.41	14621.94	32.53	18319.79	33036.32	55.77	430.50	1098.55	39.19	1434.92	3288.70	43.64	336.49	502.39	67.36
NPV_star_test2_test_	40000	4932.96	14577.32	33.79	17138.53	35101.01	49.48	429.23	1098.04	39.09	1466.38	3285.72	44.63	357.31	506.27	70.58
NPV_star_test3_test_	40000	4425.09	14915.03	29.69	17729.67	33795.97	51.63	425.02	1094.94	38.82	1609.91	3287.41	48.96	400.82	511.49	78.41

Another factor of 10 down (now at 12), only the seek time is improved (not suprising since it does not have to scan as much of the file size).

		Block IO						Character						Random
		fwrite			fread			putc			getc			seek
Machine/FS	KB	KB/sec Real	KB/sec CPU	%CPU	KB/sec Real	KB/sec CPU	%CPU	KB/sec Real	KB/sec CPU	%CPU	KB/sec Real	KB/sec CPU	%CPU	KB/sec Real	KB/sec CPU	%CPU
_star_test0_	4000	16000.00	16666.67	96.00	80000.00	80000.00	100.00	1183.43	1183.43	100.00	3418.80	3418.80	100.00	488.28	488.28	100.00
_star_test1_	4000	16666.67	16666.67	100.00	80000.00	80000.00	100.00	1183.43	1186.94	99.70	3448.28	3448.28	100.00	488.28	488.28	100.00
_star_test2_	4000	16666.67	16666.67	100.00	80000.00	80000.00	100.00	1179.94	1179.94	100.00	3448.28	3448.28	100.00	488.28	488.28	100.00
_star_test3_	4000	16000.00	16000.00	100.00	80000.00	80000.00	100.00	1183.43	1186.94	99.70	3418.80	3418.80	100.00	488.28	488.28	100.00
1PV_star_test0_test_	4000	16000.00	16666.67	96.00	100000.00	100000.00	100.00	1186.94	1186.94	100.00	3478.26	3478.26	100.00	976.56	976.56	100.00
1PV_star_test1_test_	4000	16666.67	16666.67	100.00	80000.00	80000.00	100.00	1173.02	1176.47	99.71	3478.26	3478.26	100.00	488.28	488.28	100.00
1PV_star_test2_test_	4000	16666.67	16666.67	100.00	80000.00	80000.00	100.00	1156.07	1159.42	99.71	3448.28	3448.28	100.00	488.28	488.28	100.00
1PV_star_test3_test_	4000	16000.00	16000.00	100.00	80000.00	80000.00	100.00	1055.41	1055.41	100.00	3418.80	3418.80	100.00	488.28	488.28	100.00
NPV_star_test0_test_	4000	6361.66	15085.12	42.21	49072.26	65158.73	74.80	429.94	1098.44	39.17	2252.42	3413.45	66.10	453.40	488.28	92.86
NPV_star_test1_test_	4000	6505.36	14890.29	44.03	40760.40	65190.48	65.81	439.63	1100.89	40.00	2006.54	3413.88	58.85	455.73	488.28	93.33
NPV_star_test2_test_	4000	5805.70	15026.86	38.88	42837.51	67761.91	62.02	453.47	1092.67	41.56	1892.04	3401.52	55.65	488.28	488.28	100.00
NPV_star_test3_test_	4000	5233.29	15308.98	34.48	47928.57	60476.19	79.32	440.97	1100.81	40.11	2106.61	3401.86	61.97	537.11	537.11	100.00

As before for the number of IO but this time, instead of using a 8KByte buffer (2^3) we use a 64 KByte buffer (2^6).
5 % gain in write and read mode . Comparing to the IOZone graphs, we should now be on the record size=6, file size ~ 12, bottom graphs.
Note ta this stage that test0 always seem to do better read-wise than the others. Any tuning issues ?
Also, the 1PV tests are : one thread per volume (all volume simultaneously), there are no significant differences in rates comparing to a single thread, one volume at a time. Is this expected ??

		Block IO						Character						Random
		fwrite			fread			putc			getc			seek
Machine/FS	KB	KB/sec Real	KB/sec CPU	%CPU	KB/sec Real	KB/sec CPU	%CPU	KB/sec Real	KB/sec CPU	%CPU	KB/sec Real	KB/sec CPU	%CPU	KB/sec Real	KB/sec CPU	%CPU
_star_test0_	32000	16580.31	16580.31	100.00	84210.53	84210.53	100.00	1167.03	1167.03	100.00	3433.48	3433.48	100.00	488.28	488.28	100.00
_star_test1_	32000	16842.11	16842.11	100.00	84210.53	84210.53	100.00	1188.27	1188.27	100.00	3463.20	3463.20	100.00	488.28	488.28	100.00
_star_test2_	32000	16666.67	16753.93	99.48	84210.53	84210.53	100.00	1106.88	1110.73	99.65	3440.86	3444.56	99.89	488.28	488.28	100.00
_star_test3_	32000	16666.67	16666.67	100.00	84210.53	84210.53	100.00	1140.41	1140.41	100.00	3459.46	3459.46	100.00	488.28	488.28	100.00
1PV_star_test0_test_	32000	16666.67	16753.93	99.48	86486.49	86486.49	100.00	1096.27	1114.59	98.36	3379.09	3389.83	99.68	488.28	488.28	100.00
1PV_star_test1_test_	32000	16494.85	16580.31	99.48	80000.00	80000.00	100.00	1081.81	1095.52	98.75	3389.83	3400.64	99.68	976.56	976.56	100.00
1PV_star_test2_test_	32000	16080.40	16243.65	98.99	76190.48	76190.48	100.00	1082.18	1103.83	98.04	3386.24	3418.80	99.05	488.28	488.28	100.00
1PV_star_test3_test_	32000	13733.91	15311.00	89.70	66666.67	66666.67	100.00	1099.66	1113.04	98.80	3433.48	3433.48	100.00	488.28	488.28	100.00
NPV_star_test0_test_	32000	4626.94	15152.89	30.51	19517.83	34733.29	55.94	429.30	1096.93	39.14	1482.39	3297.30	44.96	444.88	488.28	91.11
NPV_star_test1_test_	32000	4947.71	15282.74	32.36	21458.08	34283.35	63.61	427.49	1100.20	38.86	1613.10	3296.29	48.94	537.11	537.11	100.00
NPV_star_test2_test_	32000	5323.42	14787.27	35.91	18265.83	36711.84	50.54	425.83	1096.76	38.83	1542.92	3292.90	46.85	493.70	537.11	91.11
NPV_star_test3_test_	32000	5103.55	14708.89	34.68	16168.03	35475.83	45.65	428.48	1099.18	38.99	1595.63	3289.54	48.51	447.59	488.28	91.67

IOzone profiling

The profiling gives a gross performance view for several kByte buffer and several file size.
The Read/Write tests are unbuffered while FRead/FWrite tests are buffered.