LinuxQuestions.org

LinuxQuestions.org (/questions/)
-   Linux - Hardware (http://www.linuxquestions.org/questions/linux-hardware-18/)
-   -   Slow speeds while dumping backup files to LTO-3 tape device (http://www.linuxquestions.org/questions/linux-hardware-18/slow-speeds-while-dumping-backup-files-to-lto-3-tape-device-581468/)

dafart 09-01-2007 09:51 AM

Slow speeds while dumping backup files to LTO-3 tape device
 
Hello everyone. I hope someone can help me out or at least give me a hint to how to solve this issue:
I have an arch box running the latest x86_64 arch release.
I'm using the programm "dump " http://dump.sourceforge.net/to backup to my tape device (which is an HP storageworks 920)
My scsi controller is an Intel® RAID Controller SRCU42E http://www.intel.com/design/servers/RAID/srcu42e/ controller plugged into the PCIe slot of my machine. As far as I can see it's one of the few cards for scsi devices that support Ultra 320 speeds for scsi.
The problem is that with each blocksize (from 256 up to 1024) speeds are incredibly slow.
My motherboard is: ASRock 939Dual-SATA2
My question is how to debug this whole situation,
some info on my system:
Code:

# lsmod
Module                  Size  Used by
st                    40484  2
mptspi                16784  0
mptscsih              21632  1 mptspi
mptbase                55136  2 mptspi,mptscsih
sg                    31784  0
udf                    83496  1
ipv6                  302024  12
nvidia              8107320  24
ppdev                  8712  0
lp                    11720  0
snd_seq_oss            32896  0
snd_seq_midi_event      7808  1 snd_seq_oss
snd_seq                55680  4 snd_seq_oss,snd_seq_midi_event
snd_seq_device          7956  2 snd_seq_oss,snd_seq
usb_storage            91072  3
snd_pcm_oss            42144  0
snd_mixer_oss          16896  1 snd_pcm_oss
analog                10976  0
pcspkr                  3456  0
ppp_generic            27680  0
ns558                  5120  0
gameport              13712  3 analog,ns558
slhc                    6528  1 ppp_generic
rtc_cmos                8504  0
rtc_core              18060  1 rtc_cmos
rtc_lib                3456  1 rtc_core
uli526x                17044  0
parport_pc            40680  1
ehci_hcd              34956  0
psmouse                41884  0
serio_raw              6788  0
snd_intel8x0          34600  1
snd_ac97_codec        116824  1 snd_intel8x0
ac97_bus                3072  1 snd_ac97_codec
snd_pcm                82440  3 snd_pcm_oss,snd_intel8x0,snd_ac97_codec
snd_timer              22664  2 snd_seq,snd_pcm
parport                37132  3 ppdev,lp,parport_pc
k8temp                  5760  0
i2c_ali1535            7556  0
i2c_ali1563            7812  0
i2c_ali15x3            8196  0
i2c_core              23552  4 nvidia,i2c_ali1535,i2c_ali1563,i2c_ali15x3
snd                    57448  11 snd_seq_oss,snd_seq,snd_seq_device,snd_pcm_oss,snd_mixer_oss,snd_intel8x0,snd_ac97_codec,snd_pcm,snd_timer
soundcore              7968  1 snd
snd_page_alloc          8720  2 snd_intel8x0,snd_pcm
ohci_hcd              21636  0
shpchp                32924  0
pci_hotplug            31748  1 shpchp
tsdev                  8064  0
evdev                  9984  3
thermal                13200  0
processor              28952  1 thermal
fan                    4872  0
button                  7584  0
battery                9864  0
ac                      5128  0
capability              4616  0
commoncap              6528  1 capability
usbcore              132144  4 usb_storage,ehci_hcd,ohci_hcd
ext3                  131856  9
jbd                    64752  1 ext3
mbcache                8584  1 ext3
ide_disk              15744  2
ide_cd                39840  1
cdrom                  38312  1 ide_cd
osst                  56232  0
sd_mod                25088  15
alim15x3              11288  0 [permanent]
megaraid_mbox          32272  1
megaraid_mm            10680  1 megaraid_mbox
ahci                  22660  1
generic                6276  0 [permanent]
ide_core              140816  5 usb_storage,ide_disk,ide_cd,alim15x3,generic
sata_promise          12804  2
sata_uli                6404  3
libata                119440  3 ahci,sata_promise,sata_uli

Code:

lsscsi
[0:0:0:0]    disk    ATA      SAMSUNG SP2504C  VT10  /dev/sda
[2:0:3:0]    tape    HP      Ultrium 3-SCSI  D21D  /dev/st0
[3:0:0:0]    disk    ATA      SAMSUNG HD501LJ  CR10  /dev/sdb

The scsi RAID card is attached to the Tape device only, no harddiscs whatsoever are attached.

Code:

dump -b 1024 -f /dev/st0 /media/media5/AF
  DUMP: Date of this level  dump: Sat Sep  1 18:30:37 2007
  DUMP: Dumping /dev/sdb1 (/media/media5_(dir /AF)) to /dev/st0
  DUMP: Label: media5
  DUMP: Writing 1024 Kilobyte records
  DUMP: mapping (Pass I) [regular files]
  DUMP: mapping (Pass II) [directories]
  DUMP: estimated 398145918 blocks.
  DUMP: Volume 1 started with block 1 at: Sat Sep  1 18:30:38 2007
  DUMP: 0.00% done, finished in 0:00
  DUMP: dumping (Pass III) [directories]
  DUMP: dumping (Pass IV) [regular files]
  DUMP: 0.14% done at 1877 kB/s, finished in 58:49

As you can see it only writes at 1877 kB/s my tape device is also pausing a lot, so I assume it's something with the scsi raid device not working as fast as it should.

The card uses an LSI chip: http://www.lsi.com/storage_home/prod...030/index.html which uses the driver:
Fusion MPT base driver 3.04.04
Copyright (c) 1999-2007 LSI Logic Corporation
Fusion MPT SPI Host driver 3.04.04
Which I loaded manually. I also loaded the sg and the st module manually. It doesn't matter in what order I load these modules, the speed remains the same.

Anyone with any idea how to fix this? How can I make sure the controller uses it's full 320 capacity?

Any help would be great!

traene 09-02-2007 04:23 AM

What's the workload on your host? Let's have a look at top (or maybe run gkrellm). Other option: Can you try to dump to another device? Maybe just a subset of the whole data, to /tmp and maybe to /dev/null. I have no experience with tape drives. But maybe you can use hdparm or something similiar to change settings of the SCSI drive.

dafart 09-02-2007 05:59 AM

While dumping to tape using my other (uw160) controller I get much higher speeds (35700 kB/s), system load is not the issue. I can try to dump to another device (hdd) but I'm quite sure it's something with the device and the correct drivers for it..
Since the other controller gives me much highr speeds I don't think it something with the device settings.

strick1226 09-04-2007 10:35 AM

Hmm... interesting.

I have a bunch of LTO-2 and LTO-3 drives in some Netware boxes, but they're all attached to PCI-connected Adaptec 29160/39160 host adapters. The first thing the vendor told us was not to run the tape drives attached to the SCSI RAID controllers--even if on a completely separate channel.

The Asrock 939DUAL-SATA2 has a rather funky (and neat!--I have one at home--great inexpensive board) ULI chipset... I'm fairly certain no enterprise-class host adapters/RAID cards are certified to work in one.

That said, your Intel card's Ultra320 interface is way overkill for even an LTO-3 drive. Heck, Ultra160 is overkill for an LTO-3 drive w/a max transfer rate of 120 MB/s in ideal circumstances (2:1 compression, perfect relative humidity, phase of moon, etc.). PCI @ 33 MHz can provide ~266 MB/sec, so PCIX/PCIe isn't necessarily a requirement for full-speed LTO-3 operation...

If you have been able to obtain expected speeds with the Ultra160 controller you might just want to stick with that for the time being.

If you've been able to prove it's not a problem with the tape drive then I'd recommend checking out all the possible options in your intel RAID controller's BIOS options, perhaps even updating the BIOS/firmware if such an option exists. It almost sounds like there could be some kind of termination issue specific to that card...

I'm not familiar with this particular card, but there should be an option to "hard-set" the negotiation speed/rate for the LTO-3's particular ID - try capping it @ 160 and re-test etc.

Then again, I'm still unfamiliar with trying a PCIe U320 card in that board, so that may be a factor as well.

Have you attempted to contact intel/LSI support?

Good luck!


All times are GMT -5. The time now is 10:45 AM.