possible issues with Dell/Perc5 raid

[Eric Anderson]

On 05/11/07 08:42, Bill Moran wrote:
> In response to Randy Schultz <schulra at earlham.edu>:
> 
>> Hi there,
>>
>> We just purchased a Dell 860 with these specifics:
>>    - dual core pentium 915, 2.8 GHz, 800MHz FSB
>>    - 2x512 MB 533 MHz DDR2 RAM
>>    - Dell's SAS/SATA drive controller(which is actually an LSILogic)
>>    - their 5IR PCI-Express internal RAID controller
>>    - a pair of 250 GB SATA II's set up RAID 1
>>
>> After installing 6.2-RELEASE on the system I noticed, while pulling the ports
>> collection off the CDROM, that the speed to the drives started out just fine
>> but long before the ports collection was installed the transfer rate was down
>> to ~35kB/s.  Huh.  So I started paying closer attention.  I installed a couple
>> of ports and every time there was anything but miniscule data being
>> transferred to/from disk the transfer rate bottomed out.  I installed the
>> blogbench port and it confirmed the I/O as tremendously sluggish.  I've
>> attached the output of the blogbench runs.  The first is from an old 700 MHz
>> pc, Acer M25C mobo, with 1 IDE drive just for comparison.  The second data set
>> is the new Dell.  I've also attached the output of dmesg.  Everything looks
>> fine except for the mpt lines.  Would those be telling of something amiss or
>> are they innocuous barks?
>>
>> I understand that RAID 1 is not the fastest but I certainly expected it to be
>> faster than it is, esp. with hw RAID.  Are my expectations too high and this
>> is normal?
> 
> My personal server is a Dell 860 configured very similarly to yours,
> with the difference that it has a single 160G SATA drive (no RAID).
> It's also running 6.2.
> 
> Frequency = 10 secs
> Scratch dir = [/home/wmoran/blogbench]
> Spawning 3 writers...
> Spawning 1 rewriters...
> Spawning 5 commenters...
> Spawning 100 readers...
> Benchmarking for 30 iterations.
> The test will run during 5 minutes.
> 

[..snip..]
> Final score for writes:           157
> Final score for reads :          6648
> 
> For comparison, here are the scores you posted:
> 
> Frequency = 10 secs
> Scratch dir = [/var/tmp]
> Spawning 3 writers...
> Spawning 1 rewriters...
> Spawning 5 commenters...
> Spawning 100 readers...
> Benchmarking for 30 iterations.
> The test will run during 5 minutes.
> 

[..snip..]

> Final score for writes:            21
> Final score for reads :          2442
> 
> Certainly looks like that RAID card sucks.  Unfortunately, I don't know
> what those mpt0 alerts are about.
> 


Could this be a write-through vs write-back cache option difference on 
the controller?


Eric