ATTENTION: For professional reasons I'm no longer directly involved with RAID systems and for that reason am not able to follow its development directly. This page is popular and I think useful and informative, and allthough I'm still trying to maintain it I'd be very happy to pass maintenance of this page on to some interested human being. If you are interested then please contact me at tpo_hp at sourcepole.ch.
This page started with a few RAID benches. Allthough I'm still interested in RAID system performances (mainly RAID5 with Linux, because that's what we use here) I have begun including non-RAID benchmarks. The reason for it is to find out which combinations of HW/software/config yield good results as internet servers. It is clear that the combination of bonnie and bm (Byte Benchmark) as benchmarks don't permit any real conclusion on the performance of an internet server, but they give at least some numbers and can at least be used to find out what relevance which components have on performance.
If you do bonnie benches with high performance RAID or non-RAID systems on your own, I'd be happy to include them here. Please include at least the following info: processor, RAM, system version, RAID config (if ever), HD types and numbers. Note that for bonnie benchmarks it doesn't really make sense if the size of your RAM is comparable to the size of the bonnie work-file. See the note by Linas. Suggestions and comments are very wellcome: mailme at tpo_hp at sourcepole.ch.
If you are looking for further RAID reading check out Linas' RAID Solutions for Linux pages that includes more links and info. And also Russel Coker's bonnie++ page, that also includes a few benchmark results.
Unless otherwise noted, the benches were performed by myself
If you want to speed up your system, you might want to think about using hdparm.
Linas Vepstas writes a thing which make the 100MB bonnies at least
questionable:
Note that a 100MB test on a 256 MB machine is invalid: you are measuring
memory performance, not disk performance, as there will be *no* disk activity
because of (a) the Linux file system cache, and (2) Linux disk write
buffering.
For a 256M machine, I suggest at least a 1G test, although a 2 or 3 G test
would be better.
| Config |
-------Sequential Output-------- ---Sequential Input-- --Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
|
| P200, 64M, NEC DSE1700A |
tpo's 100 2906 90.2 4217 42.9 1802 43.7 2352 81.2 4742 75.6 51.5 5.4 |
| K6 233, 64M, 8 GB Maxtor 88455D8 |
IDE 100 2425 93.6 6169 24.2 2997 21.5 2454 91.8 10872 24.5 73.5 1.7 |
| with soft RAID0 o. 4x8 GB Max.88455D8 |
RAID-0 100 2252 87.5 10720 25.6 4346 21.9 2550 94.4 18052 39.7 136.8 3.1 |
| Dual PPro 200, 64M, Cheetah |
Cheetah 100 2890 95.6 11156 50.4 3847 29.1 2622 91.0 12845 31.2 253.7 6.6 |
| AMD P5 200, 16M, IBM-DCAA-34330 |
ruben's 100 4116 38.0 4349 15.0 2417 15.7 2617 22.1 4254 10.4 46.9 1.4 |
| Ultra/143 HW-RAID write-through |
RAID5 100 3205 46.6 3490 9.0 2751 9.9 5077 95.0 19314 27.8 609.7 12.3 |
| Ultra/143 HW-RAID write-back |
RAID 5 100 5922 86.1 5967 15.6 4106 14.5 5277 97.9 20341 28.5 968.3 19.3 |
| Ultra/143 fast/narrow |
ST32171N 100 6465 97.9 8610 21.1 4418 15.8 5032 97.4 22664 31.6 288.3 5.8 |
| Ultra1/143 fast/wide |
IBM34330W 100 6681 97.4 7474 18.5 3707 12.9 5248 96.9 22575 30.3 275.5 5.3 |
| PPro 200, GDT RAID5 o. 3 ST34572W |
fetty 100 6552 89.8 15329 53.6 2594 20.0 3240 36.2 62023 87.8 7630.4 99.2 |
| PPro 200, GDT RAID5 o. 3 DCAS-3433OW |
fetty-i 100 7599 97.5 22763 74.7 6673 43.5 2784 30.6 82875 67.2 6908.2 98.4 |
| same 2nd test |
fetty-i 100 7393 93.7 25898 83.5 7175 46.7 2590 29.5 130967 99.8 6535.0 88.2 |
| 2*P3 450, DAC960PD, 5 IBM DNES-318350Y 18G SCSI |
DAC960pd 100 6350 87.9 7311 5.6 4287 9.3 7385 100.0 270706 100.5 14356.2 201.0 |
| HP-PA 7100LC w. ST |
128 2320 3939 1064 1903 3493 75,8 |
| same w. IBM |
128 3485 4087 1657 2880 3462 70,8 |
| PPro-200, 64M |
128 3163 4251 1735 2623 3355 60 |
| P166, 64M |
128 4676 4373 1785 3745 4825 44 |
| AMD K5-PR166, 64M |
128 5016 3912 1788 4613 4426 43 |
| AMD K6-PR233, 128M |
256 2143 2766 1130 2188 2902 47 |
| same with RAID0 |
256 7627 10595 3695 6178 11250 94.6 |
| AMD K6-PR200, 128M |
256 3967 5233 2359 5083 5967 67 |
| DEC Alpha 21164@600MHz |
256 3395 3677 1795 4197 4286 65.4 |
| same w. IBM on SCSI |
256 3574 4410 1746 3498 4221 76.3 |
| same w. Micropolis |
256 3574 4410 1746 3498 4221 76.3 |
| same w. IBM |
256 4333 5542 2156 4643 5317 57.0 |
| Cyrix 486 120, 16M, soft RAID0 |
sfirst 800 765 90.0 1414 77.9 620 61.7 776 92.6 1169 55.5 14.0 5.1 |
| PPro 200, GDT RAID5 o. 3 DCAS-3433OW |
fetty-i 800 2554 35.8 3556 17.4 1968 15.6 6789 64.2 9870 18.9 76.2 2.2 |
| P233, 64M, 9GB IBM |
backup 800 2887 97.0 6747 31.9 3597 31.0 3135 95.2 13644 52.0 56.3 2.1 |
| PII233, 128M, RAID5 |
mail 800 2125 50.9 2974 9.4 2280 11.4 4503 92.7 10469 16.3 88.3 2.2 |
| same with 64M and 2.0.33 |
mail 800 4137 96.9 3033 7.9 2579 12.1 4481 91.6 10843 17.4 81.1 2.1 |
| P233, 128M, soft RAID0 o. 2 EID |
p-sargans 800 7037 90.8 12239 74.9 2882 30.2 4975 53.1 5576 24.7 37.1 2.0 |
| PII 266 Quant.XP32275W |
p-thusis 800 8658 86.6 8870 26.0 3208 14.4 7656 51.6 7345 10.4 36.7 0.4 |
| PII 266 soft RAID0 o. Quant.XP32275W |
p-thusis 800 9131 90.9 15203 47.5 2743 15.3 9900 70.8 10378 22.2 38.9 1.1 |
| DPT 3 disks |
DPT3x4G 1000 1914 20.0 1985 2.8 1704 6.5 5559 86.7 12857 15.6 97.1 1.8 |
| DPT 6 disks |
DPT6x4G 1000 2246 23.4 2371 3.4 1890 7.1 5610 87.3 9381 10.9 112.1 1.9 |
| soft RAID5 3 disks |
SOF3x4G 1000 7312 76.2 10908 15.5 5757 20.2 5434 86.4 14728 19.9 69.3 1.5 |
| soft RAID5 6 disks |
SOF6x4G 1000 7530 76.8 16991 32.0 7861 39.9 5763 90.7 23246 49.6 145.4 3.7 |
| soft RAID5 8 disks |
SOF8x4G 1000 7642 77.2 17649 33.0 8207 41.5 5755 90.6 22958 48.3 160.5 3.7 |
| soft RAID0 8 disks |
SOF8x4G 1000 8506 86.1 27122 54.2 11086 58.9 6077 95.9 27436 62.9 185.3 4.9 |
| F-BSD P200 RAID |
ns2a1 1200 6354 30.5 6062 8.1 4426 8.8 11214 55.0 11319 14.2 182.0 3.0 |
| F-BSD PII 400 RAID |
ns2a2 1200 6345 30.5 6068 8.1 4387 8.8 11209 55.1 11328 14.3 170.8 2.8 |
| 2*P3 450, DAC960PD, 5 IBM DNES-318350Y 18G SCSI |
DAC960pd 1800 3867 53.9 4718 4.3 2825 5.9 6116 80.5 9863 7.6 269.9 3.6 |
| 2*AMD 1600, 1024M, 8 Matrox D540X 160G IDE, RAID 5 |
Raid5 256 16707 98.0 57772 73.2 24978 32.1 17127 96.8 91138 46.2 301.6 2.8 |
| on 2G |
Raid5 2000 15892 94.1 55295 73.1 24636 33.0 16256 94.2 83222 60.6 328.6 1.9 |
| RAID 0 |
Raid0 256 18489 99.7 89231 65.7 42727 28.7 17324 97.2 115269 91.9 313.2 1.2 |
| P3 1G, 512M, 2 A39160, 8 73G ultra160 ST173404LCV SCSI |
Raid0 256 13218 99.7 90025 99.9 64460 59.3 12577 97.9 176889 85.7 409.4 4.9 |
| with 18 disks |
Raid5 256 11879 97.8 51639 98.7 19286 38.6 12040 96.4 119272 68.2 467.0 8.8 |
| RAID 0 |
Raid0 256 13192 99.6 92973 99.7 94040 95.4 12625 99.0 178365 89.1 545.1 7.6 |
| P3 1G, 512M, Fasttrack, 4x80G IBM IDE |
sirius 1024 7297 60.8 29804 38.9 19738 42.7 12301 91.6 55403 72.0 109.5 1.4 |
For comparison see The Linux Benchmarks. Klick on the name for the full bench report.
Platform tested using BYTE Benchmarks SUM AVERAGE Notes _____________________________________________________________________________ tpo Pentium S 200 Mhz 64 M RAM PCI EIDE 66.3 11.1 ruben AMD K6 200 Mhz 32 M RAM PCI EIDE 118.5 19.7 p-sarga Pentium 233 Mhz 128 M RAM PCI SCSI 236.5 39.4 backup Pentium 233 Mhz 64 M RAM PCI SCSI 258.1 43.0 fetty Pentium Pro 200 Mhz 256 M RAM PCI EIDE 316.6 52.8 mail Pentium II 233 Mhz 64 M RAM PCI SCSI 320.2 53.4 mail128MPentium II 233 Mhz 128 M RAM PCI SCSI 326.0 54.3 p-thusi Pentium II 266 Mhz 128 M RAM PCI EIDE 385.0 64.2