Raid 5 benchmark

[Indyboy2]

If Anybody running a raid 5 array please run the atto benchmark and post your results please .I was having trouble with my read /write speeds i think ive got it figured out .Preferably running on a 3ware controller but i guess any hardware controller will do for the test
http://members.home.nl/rvandes.../ATTO%20benchmark.html
just want something to base my results on with someone with a similar array setup

 

[Fardringle]

I'm afraid I can't be much help with an accurate baseline for your 3ware SATA RAID array but these are the results for the default benchmark settings on an Ultra 160 SCSI system (Dell PERC 4e/Di controller) if you want to see how your setup compares...


Size - Write/Read
.05 - 15856/17801
1.0 - 32144/33529
2.0 - 64475/67962
4.0 - 104571/123707
8.0 - 171966/208666
16.0 - 236978/303038
32.0 - 305135/369098
64.0 - 386924/488636
128.0 - 430964/478150
256.0 - 452984/505413
512.0 - 464519/524288
1024.0 - 460324/513802

 

[Madwand1]

Originally posted by: Fardringle
I'm afraid I can't be much help with an accurate baseline for your 3ware SATA RAID array but these are the results for the default benchmark settings on an Ultra 160 SCSI system (Dell PERC 4e/Di controller) if you want to see how your setup compares...


Size - Write/Read
.05 - 15856/17801
1.0 - 32144/33529
2.0 - 64475/67962
4.0 - 104571/123707
8.0 - 171966/208666
16.0 - 236978/303038
32.0 - 305135/369098
64.0 - 386924/488636
128.0 - 430964/478150
256.0 - 452984/505413
512.0 - 464519/524288
1024.0 - 460324/513802

Those are incredible figures. How many hard drives were in the array? Which model drives?

Are you sure that those are realistic figures not saturating the test size with cache? Did you use the 256 MB test size / version which can be found through here?

http://kb.ciprico.com/lore/article.php?id=268

 

[Fardringle]

I'm not near the server right now so I can't look at the drives to see the exact model number, but there are four Maxtor U320 73GB drives in the server.

It's entirely possible that the whole test ran within the cache of the drives. As I said, I just ran the default benchmark settings on the version that Indyboy posted. I'll run the 256MB version you posted in a little while (when nobody is using the server during lunch) and see what it says.

 

[Indyboy2]

Test results
size,write/read
0.5 /11897 /8086
1.0 /24313 /15950
2.0 /47710 /32899
4.0 /84896 /63176
8.0 /150339 /103038
16.0 /212795 /127860
32.0 /271177/ 134676
64.0 /38338 /142671
128.0 /307429 /138018
256.0 /30453 /140640
512.0 /55421 /244580
1024.0 /72853 /264503
Settings transfer size 0.5 to 1024 kb
Total Length 256kb
Seems like the write speeds are slow?

 

[Fardringle]

Unless you're only using the cache on the hard drives, writing is always slower than reading, particularly on a striped array. It is rather interesting that you had such large spikes in write speed at 8, 16, 32, and 64KB transfer sizes compared to the rest of your results.

These are my results for the 256MB test. As you can see, reading is still very fast as it should be, but the write results are much lower than my first run because the test is no longer just running within the cache of the drives and actually has to wait for the array to write the data to the physical disks before it can proceed.


0.5 - 16343/18294
1.0 - 31409/34322
2.0 - 57189/68132
4.0 - 67108/120119
8.0 - 64079/201869
16.0 - 64863/290110
32.0 - 67232/379785
64.0 - 55733/495103
128.0 - 41646/472969
256.0 - 45497/521233
512.0 - 52123/532874
1024.0 - 48020/528936
2048.0 - 48193/531555
4096.0 - 44442/494659
8192.0 - 50268/472597