linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* WD Raptor/SATA with RAID0 way to slow
@ 2003-10-31  2:49 Julien Oster
  2003-10-31 17:34 ` Torrey Hoffman
  0 siblings, 1 reply; 3+ messages in thread
From: Julien Oster @ 2003-10-31  2:49 UTC (permalink / raw)
  To: linux-kernel


Hello,

I recently purchased an Asus A7N8X mainboard with the Silicon Imaging
SATA controller online and two Western Digital Raptor SATA harddrives
with 10krpm.

Those harddisks are supposed to be "really fast", but I don't really
get the performance out of them. In fact, I get much less performance
than with my "standard consumer" IBM DeskStars.

Besides the WD Raptors, I have two IBM DeskStars (normal PATA). The
two IBM deskstars and the two WD Raptors are bundled to linux
softraids with RAID level 0 each.

I use the same hdparm command line for all drives during boot, which
is:

hdparm -c 1 -m 16 -S 241 -u 1 -k 1 -K 1

I first used a simple hdparm -t on the harddisks individually to see
how they perform alone. Here's the values I got for one of the IBMs:

IBM DeskStar
 Timing buffered disk reads:  64 MB in  1.41 seconds = 45.39 MB/sec

And now for the Raptors:
WD Raptor
 Timing buffered disk reads:  64 MB in  1.24 seconds = 51.61 MB/sec

Well, at least it's faster. But the next test is really annoying me. I
did the same hdparm -t on the RAID0 containing the IBMs (/dev/md5) and
the one containing the Raptors (/dev/md6) and got the following
result:

/dev/md5:
 Timing buffered disk reads:  64 MB in  0.71 seconds = 90.14 MB/sec

/dev/md6:
 Timing buffered disk reads:  64 MB in  1.06 seconds = 60.38 MB/sec

For the IBMs, the value is fine, it's nearly double the throughput
than with a single drive. But for the Raptor RAID, it's even less than
for the IBM RAID! It doesn't even come close to doubling the
throughput, which is the fact for the IBM drives.

Another quick test with bonnie++ showed that the Raptor RAID really is
slower then the IBM RAID. Not much, but it is - and that despite the
fact that it should be much faster. The Raptors have a very low avg
read seek time of 4.5ms while the IBMs do have 9.8ms, so I expected at
least the filesystem test with bonnie++ to show some performance
improvement, when doing things like creating and deleting files, but
absolutely no test didn't gave a lower performance for the Raptors.

Any idea of what could be wrong? This really annoys me, because I
purchased the smaller Raptors just because I need something as fast as
possible, due to the extensive data amounts I'm working it (and SCSI
not being an option for my development machine at home).

Thanks in advance,
Julien

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: WD Raptor/SATA with RAID0 way to slow
  2003-10-31  2:49 WD Raptor/SATA with RAID0 way to slow Julien Oster
@ 2003-10-31 17:34 ` Torrey Hoffman
  2003-10-31 19:08   ` Julien Oster
  0 siblings, 1 reply; 3+ messages in thread
From: Torrey Hoffman @ 2003-10-31 17:34 UTC (permalink / raw)
  To: Julien Oster; +Cc: Linux Kernel

On Thu, 2003-10-30 at 18:49, Julien Oster wrote:
> Hello,
> 
> I recently purchased an Asus A7N8X mainboard with the Silicon Imaging
> SATA controller online and two Western Digital Raptor SATA harddrives
> with 10krpm.
> 
> Those harddisks are supposed to be "really fast", but I don't really
> get the performance out of them. In fact, I get much less performance
> than with my "standard consumer" IBM DeskStars.

What version of the Linux kernel are you using?  What does the kernel
log say about how it is detecting and configuring your drives?


-- 
Torrey Hoffman <thoffman@arnor.net>


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: WD Raptor/SATA with RAID0 way to slow
  2003-10-31 17:34 ` Torrey Hoffman
@ 2003-10-31 19:08   ` Julien Oster
  0 siblings, 0 replies; 3+ messages in thread
From: Julien Oster @ 2003-10-31 19:08 UTC (permalink / raw)
  To: Torrey Hoffman; +Cc: Julien Oster, Linux Kernel

Torrey Hoffman <thoffman@arnor.net> writes:

Hello,

>> Those harddisks are supposed to be "really fast", but I don't really
>> get the performance out of them. In fact, I get much less performance
>> than with my "standard consumer" IBM DeskStars.

> What version of the Linux kernel are you using?

Hmpf. The part of mentioning the kernel version in use is so obviously
elementary that I'm always missing it :)

It's 2.4.22-ac4. I'm using this kernel, because I had no real luck
with other kernels on my mainboard (the well-known A7N8X lockups that
are being discussed here). This one seems to work well.

> What does the kernel log say about how it is detecting and
> configuring your drives?

It's hard with 2.4.22-ac4 to get a good kernel log, since because
ACPI, APIC and the not less than seven softraids the log buffer is a
bit too small, but I managed to get everything from bootup except the
first lines.

I'll not only paste the relevant lines for the WD Raptors, but also
for the IBM harddrives, maybe there's something to compare or the
like.

The PATA controller says the following:

NFORCE2: IDE controller at PCI slot 00:09.0
NFORCE2: chipset revision 162
NFORCE2: not 100% native mode: will probe irqs later
AMD_IDE: Bios didn't set cable bits correctly. Enabling workaround.
AMD_IDE: Bios didn't set cable bits correctly. Enabling workaround.
AMD_IDE: nVidia Corporation nForce2 IDE (rev a2) UDMA133 controller on pci00:09.
0
    ide0: BM-DMA at 0xf000-0xf007, BIOS settings: hda:DMA, hdb:DMA
    ide1: BM-DMA at 0xf008-0xf00f, BIOS settings: hdc:DMA, hdd:DMA

No problem with it so far, I get really good results with my PATA
drives, it seems to be unsignificantly faster than my old board. I got
lockups on high load on the DeskStars with various other kernels, but
as I said before I managed to solve this problem by using the latest
-ac4.

Now the Silicon Image SATA controller:

SiI3112 Serial ATA: IDE controller at PCI slot 01:0b.0
SiI3112 Serial ATA: chipset revision 2
SiI3112 Serial ATA: not 100% native mode: will probe irqs later
    ide2: MMIO-DMA , BIOS settings: hde:pio, hdf:pio
    ide3: MMIO-DMA , BIOS settings: hdg:pio, hdh:pio

That looks different. What is MMIO-DMA, why not BM-DMA? And the BIOS
settings are programmed IO, not DMA... don't know if that's relevant
at all, I'm doing hdparm.

next comes the IBM harddisks plus CD-RW:

hda: C/H/S=39420/16/255 from BIOS ignored
hda: IC35L080AVVA07-0, ATA DISK drive
blk: queue c044c0c0, I/O limit 4095Mb (mask 0xffffffff)
hdc: IC35L080AVVA07-0, ATA DISK drive
hdd: SAMSUNG CD-R/RW SW-408B, ATAPI CD/DVD-ROM drive
blk: queue c044c51c, I/O limit 4095Mb (mask 0xffffffff)

nice.

WD Raptors:

hde: WDC WD360GD-00FNA0, ATA DISK drive
blk: queue c044c978, I/O limit 4095Mb (mask 0xffffffff)
hdg: WDC WD360GD-00FNA0, ATA DISK drive
blk: queue c044cdd4, I/O limit 4095Mb (mask 0xffffffff)

fine.

ide stuff:

ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
ide1 at 0x170-0x177,0x376 on irq 15
ide2 at 0xf8808080-0xf8808087,0xf880808a on irq 18
ide3 at 0xf88080c0-0xf88080c7,0xf88080ca on irq 18

ide0 and ide1 are PATA. both on different interrupts. ide2 and ide3
share the same... could THAT be the pitfall I'm stepping into, the
reason why one Raptor alone is faster than one DeskStar alone, but the
WD RAID is slower than the IBM RAID?

BTW, it *really* is slower. No test, wether it's hdparm -t, bonnie++
or my own tests with various applications, does not give a bad result
for the Raptor RAID compared to the DeskStar RAID.

Well, I just have no idea. IDE stuff is something I have no clue of. I
tried different things, with or without APIC, ACPI, hdparm... most of
those experiments do nothing at all, only with hdparm I managed to get
it even slower.

And to make it complete, here's the output of hdparm -i for one
DeskStar (/dev/hda) and one Raptor (/dev/hde):

/dev/hda:

 Model=IC35L080AVVA07-0, FwRev=VA4OA52A, SerialNo=VNC402A4CMNT6A
 Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs }
 RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=52
 BuffType=DualPortCache, BuffSize=1863kB, MaxMultSect=16, MultSect=16
 CurCHS=4047/16/255, CurSects=16511760, LBA=yes, LBAsects=160836480
 IORDY=on/off, tPIO={min:240,w/IORDY:120}, tDMA={min:120,rec:120}
 PIO modes:  pio0 pio1 pio2 pio3 pio4 
 DMA modes:  mdma0 mdma1 mdma2 
 UDMA modes: udma0 udma1 udma2 udma3 udma4 *udma5 
 AdvancedPM=yes: disabled (255) WriteCache=enabled
 Drive conforms to: ATA/ATAPI-5 T13 1321D revision 1:  2 3 4 5

/dev/hde:

 Model=WDC WD360GD-00FNA0, FwRev=35.06K35, SerialNo=WD-WMAH91116490
 Config={ HardSect NotMFM HdSw>15uSec SpinMotCtl Fixed DTR>5Mbs FmtGapReq }
 RawCHS=16383/16/63, TrkSize=57600, SectSize=600, ECCbytes=74
 BuffType=DualPortCache, BuffSize=8192kB, MaxMultSect=16, MultSect=16
 CurCHS=65535/1/63, CurSects=4128705, LBA=yes, LBAsects=72303840
 IORDY=on/off, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120}
 PIO modes:  pio0 pio1 pio2 pio3 pio4 
 DMA modes:  mdma0 mdma1 mdma2 
 UDMA modes: udma0 udma1 udma2 udma3 udma4 udma5 *udma6 
 AdvancedPM=no WriteCache=enabled
 Drive conforms to: device does not report version:  1 2 3 4 5 6

I would really appreciate some help there...

Regards,
Julien

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2003-10-31 19:08 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-10-31  2:49 WD Raptor/SATA with RAID0 way to slow Julien Oster
2003-10-31 17:34 ` Torrey Hoffman
2003-10-31 19:08   ` Julien Oster

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).