Tape drives

SGI hardware problems, solutions, tips, hacks, etc.
Forum rules
Any posts concerning pirated software or offering to buy/sell/trade commercial software are subject to removal.

Best tape archiver tool?

tar
9
56%
cpio
1
6%
xfsdump
6
38%
 
Total votes: 16

User avatar
mia
Posts: 1055
Joined: Wed Feb 19, 2003 1:54 pm

Re: Tape drives

Unread postby mia » Wed Nov 21, 2012 9:27 pm

It is an impressive setup... how much disk space do you have, and how many spindles total?
:Onyx2:

User avatar
hhoffman
Posts: 91
Joined: Fri Apr 01, 2011 7:45 am
Location: Berlin, Germany

Re: Tape drives

Unread postby hhoffman » Thu Nov 22, 2012 12:17 am

The array was actually sold this year.

User avatar
mia
Posts: 1055
Joined: Wed Feb 19, 2003 1:54 pm

Re: Tape drives

Unread postby mia » Thu Nov 22, 2012 2:12 pm

That's really an impressive flame setup. Anyway, Hhoffman, your benchmarks have shown that I'm doing something wrong, and there's room for improvement, since we have a fairly comparable hardware, would you know why I'm not getting the same throughput as yours?
:Onyx2:

User avatar
hhoffman
Posts: 91
Joined: Fri Apr 01, 2011 7:45 am
Location: Berlin, Germany

Re: Tape drives

Unread postby hhoffman » Fri Nov 23, 2012 12:39 am

maybe it depends on your configuration and setup. What irix version are you running?

User avatar
mia
Posts: 1055
Joined: Wed Feb 19, 2003 1:54 pm

Re: Tape drives

Unread postby mia » Fri Nov 23, 2012 12:59 am

hhoffman wrote:maybe it depends on your configuration and setup. What irix version are you running?


6.5.30
:Onyx2:

User avatar
hhoffman
Posts: 91
Joined: Fri Apr 01, 2011 7:45 am
Location: Berlin, Germany

Re: Tape drives

Unread postby hhoffman » Fri Nov 23, 2012 2:53 am

I'm using 6.5.28f. Maybe that is a reason ...?

Also I have upgraded to 8GB RAM a few months ago.
What is your setup?

User avatar
recondas
Moderator
Moderator
Posts: 5441
Joined: Sun Jun 06, 2004 5:55 pm
Location: NC - USA

Re: Tape drives

Unread postby recondas » Sun Mar 02, 2014 7:39 am

Another data point for mia's excellent compilation of tape drive performance, using a Quantum TC-L42AN LTO-4 SAS drive.

The test file used was located on a single spindle (an internal 15k 73G Seagate ST373454LC) connected to the IO9 SCSI controller in an 8P O350/V12 running 6.5.30:

Code: Select all

diskperf -W -D -r 4k -m 4m testfile
#---------------------------------------------------------
# Disk Performance Test Results Generated By Diskperf V1.2
#
# Test name     : Unspecified
# Test date     : Sun Mar  2 09:31:16 2014
# Test machine  : IRIX64 electraglide 6.5 07202013 IP35
# Test type     : XFS data subvolume
# Test path     : testfile
# Request sizes : min=4096 max=4194304
# Parameters    : direct=1 time=10 scale=1.000 delay=0.000
# XFS file size : 4294967296 bytes
#---------------------------------------------------------
# req_size  fwd_wt  fwd_rd  bwd_wt  bwd_rd  rnd_wt  rnd_rd
#  (bytes)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)
#---------------------------------------------------------
       4096    1.01   24.70    1.03    1.04    1.00    1.13
       8192    2.00   42.88    2.09    2.09    1.93    2.24
      16384    3.93   67.88    4.28    4.28    3.83    4.38
      32768    7.54   92.00    8.96    8.96    7.28    8.32
      65536   13.97   95.66   19.77   19.82   13.59   15.02
     131072   24.41   95.85   49.65   49.59   23.49   25.66
     262144   38.93   95.81   50.24   50.23   37.64   38.78
     524288   55.26   95.85   63.86   50.32   53.20   54.35
    1048576   70.12   95.78   81.12   62.31   67.31   62.83
    2097152   80.91   95.77   81.64   81.66   77.93   77.90
    4194304   87.72   95.76   85.29   81.63   84.07   82.97
The (external) tape drive was connected to an LSI SAS1068 controller:

Code: Select all

Integral SCSI controller 3: Version SAS/SATA LS1068
  Tape drive: unit 4 on SCSI controller 3: Linear Tape-Open

Code: Select all

# mt status
        Controller: SCSI
        Device: QUANTUM: ULTRIUM 4       21704506
        Status: 0x20242
        Drive type: Linear Tape-Open
        Media : READY, writable, at BOT

# mt blksize

 Recommended tape I/O size: 262144 bytes (512 512-byte blocks)
 Minimum block size: 1 byte(s)
 Maximum block size: 16777215 bytes
 Current block size: 4096 byte(s)

Code: Select all

mkfile -nv 46936088000 testfile.deleteme

Code: Select all

 timex tar cvKbf 4096 /dev/rmt/tps3d4 testfile.deleteme
Warning: Inclusion of file -> testfile.deleteme will create a non-portable archive
a testfile.deleteme 91672047 blocks

real    10:09.56
user        0.25
sys      4:15.50
Or roughly 73.37 MB/s (the test file was 46936088000 bytes/44761.7 megabytes, divided by 10m10s/610 seconds clock time). The test file was on a single U160 disk, so it might be possible to increase performance if the test file was read from an array of disks with higher throughput.

I used the same tape drive definition file in /master.d/scsi as hhoffman with the device info modified to match the Quantum drive used:

Code: Select all

/* Quantum LTO-4 HH SAS */
{ TPLTO, TPLTO, 7, 9, "QUANTUM", "ULTRIUM 4", 0, 0, {0},
MTCAN_BSF|MTCAN_BSR|MTCAN_APPEND|MTCAN_SETMK|MTCAN_PART|MTCAN_PREV|
MTCAN_SYNC|MTCAN_SPEOD|MTCAN_CHKRDY|MTCAN_VAR|MTCAN_SETSZ|MTCANT_IMM|MTCAN_BUFFM|
MTCAN_SILI|MTCAN_AUDIO|MTCAN_SEEK|MTCAN_CHTYPEANY|MTCAN_COMPRESS,
40, 5*60, 20*60, 20*60, 3*3600, 4096, 512*512,
tpsc_default_dens_count, tpsc_default_hwg_dens_names,
tpsc_default_alias_dens_names,
  {0}, 0, 0, 0,
   0, (u_char *)0 },
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************

User avatar
recondas
Moderator
Moderator
Posts: 5441
Joined: Sun Jun 06, 2004 5:55 pm
Location: NC - USA

Re: Tape drives

Unread postby recondas » Thu Mar 06, 2014 10:13 am

Another LTO-4 drive with an SAS interface, this time an IBM Half High LTO Gen 4 SAS Tape Drive (got lucky and got both of 'em in in the same 2U enclosure for a I-can't-believe-no-one-else-bid price).

Same test system and parameters as with the Quantum LTO-4 Half-height Model TC-L42AN SAS Tape Drive tested in the previous post.

Code: Select all

# mt status
        Controller: SCSI
        Device: IBM: ULTRIUM-HH4     B6W1
        Status: 0x20242
        Drive type: Linear Tape-Open
        Media : READY, writable, at BOT

Code: Select all

# timex tar cvKbf 4096 /dev/rmt/tps3d4v testfile.deleteme
Warning: Inclusion of file -> testfile.deleteme will create a non-portable archive
a testfile.deleteme 91672047 blocks

real     9:37.90
user        0.34
sys      4:07.86
Or approximately 78.80 MB/s (46936088000 bytes/44761.7 megabytes, divided by 9m38s/578 seconds clock time). Slightly higher than the Quantum SAS, though the limiting factor in either case is likely the throughput of single U160 hard drive. I have an external SATA hard drive enclosure that holds up to eight disks. Strictly consumer grade stuff when it comes to things like redundant data storage or even redundant power supplies, but it's desktop quiet (magnitudes quieter than the typical enterprise storage array) and relatively inexpensive to purchase / run. Did I mention it's *quiet* :D.

So I'd like to test the SAS tape drives against desktop SATA array, but first I'll have to resolve a few SAS controller issues.

The SAS controller I'm using is an LSI SAS3800X with two external SFF-8470 ports (the tape drives are both in a 2U rack enclosure, each drive has an external SFF-8088/miniSAS connection).

Initially had some issues with the SAS3800X until I isolated it as the only PCI device in the slave O350. If I connect to both ports on the SAS3800X it results in fatal PCI_Bridge errors. After doing some research I'm pretty sure I can resolve the won't pci-bus cohabitate / pci bridge errors by updating the firmware on the SAS3800X. Unfortunately I can't flash the board from IRIX. lsflash just hangs there without returning the prompt. For anyone else who might consider trying this, I'll mention I was running lsflash against firmware d/l'd from LSI - a process *not* recommended unless you're willing to accept the risk of bricking the controller. I chose to try it because while man lsflash does mention using lsflash with LSI's 1064/1068 controllers, SGI doesn't include 1064/1068 firmware (in /usr/firmware/ls).

So I'll have to temporarily put the project on standby until I get access to a PC with 3.3V PCI-X slots.
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************

User avatar
mia
Posts: 1055
Joined: Wed Feb 19, 2003 1:54 pm

Re: Tape drives

Unread postby mia » Thu Mar 06, 2014 10:42 am

That's some great work you did recondas.
:Onyx2:

User avatar
foetz
Moderator
Moderator
Posts: 6593
Joined: Mon Apr 14, 2003 4:34 am
Contact:

Re: Tape drives

Unread postby foetz » Thu Mar 06, 2014 4:14 pm

the poll lacks bru :P

User avatar
recondas
Moderator
Moderator
Posts: 5441
Joined: Sun Jun 06, 2004 5:55 pm
Location: NC - USA

Re: Tape drives

Unread postby recondas » Thu Mar 06, 2014 4:29 pm

mia wrote:That's some great work you did recondas.
Thanks! I was hoping you'd drop by. I've run the tests several times, the results are pretty consistent.



foetz wrote:the poll lacks bru :P

Everything looks better after a couple of brus. :D
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************

User avatar
recondas
Moderator
Moderator
Posts: 5441
Joined: Sun Jun 06, 2004 5:55 pm
Location: NC - USA

Re: Tape drives

Unread postby recondas » Fri Mar 07, 2014 6:56 pm

Got the firmware on the LSI controller reflashed. That seems to have resolved the issues with PCI_Bridge errors and conflicts with other PCI devices.

With the firmware issues resolved the LSI SAS controller was installed in a Tezro to see if the differences in PCI bus speed had an effect on performance (the PCI busses in an O350 limit PCI devices to a maximum of 100MHz, while the Tezro doesn't limit 133MHz PCI devices):

Code: Select all

# l1cmd pci
                                   
Bus Slot Stat    Power Mode/Speed 
--- ---- ------- ----- -----------
  1    1 0x80 01   15W PCI   66MHz 
  2    1 0x00 0f  none PCI   33MHz
  2    2 0x00 00  7.5W PCI   33MHz 
  2    3 0x00 01   15W PCI   33MHz
  3    1 0x80 0f  none PCI   66MHz
  3    2 0x80 0f  none PCI   66MHz
  4    1 0x00 0c  7.5W PCIX 133MHz   <----- The LSI SAS3800X
  4    2 0x00 0f  none PCIX 133MHz
I used identical test parameters, the hard drive used this time was a 146 GB Maxtor Atlas 15K2 connected to the IO9 in a Tezro:

Code: Select all

# mkfile 4g testfile
# diskperf -W -D -r 4k -m 4m testfile
#---------------------------------------------------------
# Disk Performance Test Results Generated By Diskperf V1.2
#
# Test name     : Unspecified
# Test date     : Fri Mar  7 20:54:16 2014
# Test machine  : IRIX64 tezro 6.5 07202013 IP35
# Test type     : XFS data subvolume
# Test path     : testfile
# Request sizes : min=4096 max=4194304
# Parameters    : direct=1 time=10 scale=1.000 delay=0.000
# XFS file size : 4294967296 bytes
#---------------------------------------------------------
# req_size  fwd_wt  fwd_rd  bwd_wt  bwd_rd  rnd_wt  rnd_rd
#  (bytes)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)
#---------------------------------------------------------
       4096    1.01   22.28    1.03   12.13    1.03    1.15
       8192    2.00   39.23    2.10   19.50    2.03    2.28
      16384    3.90   61.43    4.29   21.49    3.97    4.35
      32768    7.42   86.02    9.00   30.68    7.52    8.16
      65536   13.80   96.79   19.92   35.13   13.51   14.65
     131072   30.35   96.65   26.04   39.15   23.04   24.12
     262144   45.50   96.90   47.72   52.76   36.36   38.03
     524288   51.05   52.79   53.91   53.31   49.45   52.25
    1048576   64.09   66.89   65.82   68.96   61.73   64.94
    2097152   73.34   78.50   75.01   77.91   71.34   73.37
    4194304   80.39   85.54   80.85   85.48   78.35   84.25
The tape drive tested was the IBM Half High LTO Gen 4 SAS Tape Drive

Code: Select all

# mt status
        Controller: SCSI
        Device: IBM: ULTRIUM-HH4     B6W1
        Status: 0x20242
        Drive type: Linear Tape-Open
        Media : READY, writable, at BOT

Code: Select all

 mkfile -nv 46936088000 testfile.deleteme
testfile.deleteme 46936088000 bytes

Code: Select all

# timex tar cvKbf 4096 /dev/rmt/tps8d0v testfile.deleteme
Warning: Inclusion of file -> testfile.deleteme will create a non-portable archive
a testfile.deleteme 91672047 blocks

real     8:06.70
user        0.28
sys      3:57.85
Not sure if the increase in performance is due to the firmware, the faster PCI bus, or a combination of both, but the difference was noticeable: 91.91 MB/s (46936088000 bytes/44761.7 megabytes, divided by 8m7s/487 seconds clock time).
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************

User avatar
hamei
Posts: 10435
Joined: Tue Feb 24, 2004 4:10 pm
Location: over the rainbow

Re: Tape drives

Unread postby hamei » Fri Mar 07, 2014 9:03 pm

recondas wrote:The tape drive tested was the IBM Half High LTO Gen 4 SAS Tape Drive

Code: Select all

# mt status
        Controller: SCSI
        Device: IBM: ULTRIUM-HH4     B6W1
        Status: 0x20242
        Drive type: Linear Tape-Open
        Media : READY, writable, at BOT

/me slinks back into trailer under the bridge down by the river, giving 2mb/hr Sony SDT-9000 a little pat on the transport drive : "It's okay, little feller. He can't do audio !"

You could record the 1812 Overture in 13 seconds on that thing :( pah rah da ba da ba da rump da bum ...
I spent a fortune on booze, birds, and fast cars ... the rest I just squandered

User avatar
recondas
Moderator
Moderator
Posts: 5441
Joined: Sun Jun 06, 2004 5:55 pm
Location: NC - USA

Re: Tape drives

Unread postby recondas » Sat Mar 08, 2014 8:20 am

hamei wrote:"It's okay, little feller. He can't do audio !"
Haven't been masochistic enough to try it, but it is interesting to note that a couple of the LTO definition files include MTCAN_AUDIO.

(mtcan_audio? take a peek inside /usr/include/sys/mtio.h for the mt gospel according to irix).
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************


Return to “SGI: Hardware”

Who is online

Users browsing this forum: Yahoo [Bot] and 0 guests