3

Have just had a new HP DL380 G6 server set up.

This has a SmartArray P410i controller and seven 2.5" 146Gb SAS 10k RPM disks, configured as three RAID1 arrays and a floating hot spare.

Problem: I did a disk performance test on the drives using DiskMark and got a really LOW performance figure: 167MB/sec read, 74MB/sec write.

Compare this with a two-year old Dell PERC5i (our current server) - this has a RAID1 array of two 66Gb SAS 10K drives - DiskMark rates this at 706MB/sec read and 622MB/sec write.

I've searched for drivers and bios on HP site not found any (except for drive spec changes).. Firmware on controller reports 1.62 and Windows x64 driver is version 6.14.0.64

Any suggestions? or is the HP SmartArray controller a dog with fleas?

3
  • 1
    Have you got any cache on the controller? Depending on which model DL380 G6 you got (entry/base/performance/high efficiency) you may or may not have got any with the box and may need to purchase it seperately. Having the 512MB BBWC cache can REALLY improve the statistics. Aug 11, 2009 at 12:12
  • Controller reports 256MB cache but no battery.
    – Quango
    Aug 11, 2009 at 12:50
  • 1
    Your Dell numbers cannot be accurate, no rotational disk drive out there has a transfer speed within an order of magnitude of those figures. It looks like the Dell driver is doing some software caching or is making better use of the hardware cache. Try setting the cache ratio on your HP controller to 100% write.
    – bta
    Feb 22, 2010 at 18:50

2 Answers 2

1

I know that G6 box very well and certainly the read performance looks fine, they are after all just two disks R1's so 167MBps seems ok as that drive does about 90Mbps max each in real use. The write speed is a little low I have to say, how's the cache split?

Oh and there's no way in hell any two spinning disks will give you a real 706/622 - maybe it's just hitting the cache? You'd struggle to get a pair of SSDs to that level.

And no, smartarrays are wonderful things - I think you've just got your dell speeds wrong and perhaps the write performance might need a bit of tweaking - try Brent's sqlio guide and come back to us/me (CLICKY)

7
  • Thanks for the pointer. I realise cache is probably playing a part, but it then suggests the cache on the DELL is way better than the HP. I downloaded SQLIO, and set the file size to 1024MB so it's bigger than the cache, and ran basic single thread read test. HP: 5144 IOs/sec ~ 10MB/sec DELL: 15059 IOs/sec ~ 29MB/sec So SQLIO seems to confirm the result - the SmartArray is not as fast
    – Quango
    Aug 11, 2009 at 12:57
  • Hmm.. the controller says it's on PCI interface - I don't have physical access to server - is it on the motherboard or is is a card? if a card, has the installer plugged a PCI-X card into a PCI slot??
    – Quango
    Aug 11, 2009 at 13:00
  • Ah just seen the specs, its PCI-Express so cannot be in the wrong slot. I am going to raise a fault with HP.
    – Quango
    Aug 11, 2009 at 13:03
  • That machine is PCI-express through and through, you have to go out of your way to get PCI/PCI-X, also they aren't plug compatible. I genuinely don't think you have a problem to be honest, might rethink the raid setup however, not quite sure why you've got them as you have.
    – Chopper3
    Aug 11, 2009 at 14:12
  • Well i tried a different approach and set the testfile to 500MB on DiskMark - the HP had the same sort of stats as for 100MB but the DELL suddenly had really LOW stats.. a tenth of the performance of the HP.. so it certainly looks like cacheing is a factor. I am going to ignore all the benchmarks, load a copy of the database and see what the realworld SQL performance is like. Thanks for the help!
    – Quango
    Aug 11, 2009 at 16:23
0

Did you check to see if the cache is enabled?

I seem to recall that without a battery you have to manually enable it (had a similar issue once).

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .