DPT SmartRAID + Micropolis performance problems

DPT SmartRAID + Micropolis performance problems

Post by Daniel Jimen » Sun, 05 Jan 1997 04:00:00



I just installed a new DPT SmartRAID 'PM3224' RAID controller with 4MB
cache on my Linux system, and have it controlling two Micropolis 3243W Wide
SCSI 4.1GB internal drives (firmware 1128RV), no other SCSI devices on the
system.  I expected good performance from the controller, better than my
el-cheapo NCR53c825 card.  This isn't turning out to be the case :-(

I'm running Linux kernel version 2.0.27 on a Pentium 150

Here's what hdparm says using the '825 controller on the drives on an
otherwise unloaded system:
/dev/sda:
 Timing buffer-cache reads:   32 MB in  1.01 seconds =31.68 MB/sec
 Timing buffered disk reads:  16 MB in  2.59 seconds = 6.18 MB/sec
 Estimating raw driver speed: 16 MB in  2.08 seconds = 7.67 MB/sec

/dev/sdb:
 Timing buffer-cache reads:   32 MB in  1.00 seconds =32.00 MB/sec
 Timing buffered disk reads:  16 MB in  2.62 seconds = 6.11 MB/sec
 Estimating raw driver speed: 16 MB in  2.12 seconds = 7.55 MB/sec

Here's what it says with the DPT controller:
/dev/sda:
 Timing buffer-cache reads:   32 MB in  1.00 seconds =32.00 MB/sec
 Timing buffered disk reads:  16 MB in  3.71 seconds = 4.31 MB/sec
 Estimating raw driver speed: 16 MB in  3.21 seconds = 4.98 MB/sec

/dev/sdb:
 Timing buffer-cache reads:   32 MB in  1.01 seconds =31.68 MB/sec
 Timing buffered disk reads:  16 MB in  3.69 seconds = 4.34 MB/sec
 Estimating raw driver speed: 16 MB in  3.19 seconds = 5.02 MB/sec

Clearly, the '825 card is pulling data out 2.5MB/s faster than the DPT
controller (other tests, e.g., time dd if=/dev/sda... show the same thing).

When I combine the drives into a RAID-0 array with the DPT card (the whole
reason I bought the card, so I could get fast RAID), performance on the array
stays about the same as the individual drives, i.e., ~5 MB/s, maybe a little
faster.

Some relevant? info from /proc/scsi/eata_dma:
scsi1 : HBA PM3224A  
Firmware revision: v07G.0
Hardware Configuration:
IRQ: 10, level triggered
DMA: BUSMASTER
CPU: MC68030 40MHz
Base IO : 0xef90
Host Bus: PCI
SCSI Bus: WIDE Speed: 10MB/sec.
                      ^^^^^^^^
I thought fast wide scsi-2 should to say 20MB/sec here, i.e., six*
bits coming across at 10MHz?

So my question is, does anyone know how I can make things faster with my DPT
controller?  I've tried everything I can find documented, I'm expecting there
to be some secret incantation that I am missing.  The README file that came
with the controller said to upgrade the Micropolis firmware if the version
number is below some N, but mine are >N.

Thanks...
--

"I've so much music in my head" -- Maurice Ravel, shortly before his death.
"                             " -- John Cage

 
 
 

1. Striping (RAID-0) Performance Testing: vinum vs. DPT SmartRAID IV...

Folks,

    I don't know if anyone here cares about this or not, but I've finally
managed to sit down and correlate some performance testing I had
previously done.  I had sent these numbers to Greg Lehey (author of vinum)
back when I generated them, but I had never sat down and put everything
into charts.

    See <http://www.shub-internet.org/brad/FreeBSD/vinum.html> for the details.

--

Are you a USENET news server administrator outside of North America or
Western Europe?  Would you like a newsfeed from a site in the Freenix
Top 100?  If so, please contact me via private e-mail for details.

2. Is a boot problem in r3.1 fixed in the next release?

3. Potential problems with DPT SmartRAID V in BSDI and FreeBSD environ...

4. [Fwd: Question about shell scripting...]

5. EATA driver with DPT SmartRAID V

6. quotas on a filesystem

7. Any experiences with DPT "Decade" PM1554U2 SmartRAID

8. Performance with 3Com 595?

9. DPT SmartRaid and Everest

10. Adaptec 2100S DPT I2O SmartRAID and 2.2/2.4

11. DPT SmartRAID V/LVD/Linux Woes

12. DPT PM3224 SmartRaid and 1.2.13

13. DPT SmartRAID & eata_dma