LTO tape drive speed

[Arman Mahboobi]

Hi,

We have recently purchased and installed a LTO 230 tape drive into a N class server (hp-ux 11.00). Backups and restores to the drive using tar is ok, however, Sybase backups are slow and restores are extreemly slow (just over 2 GB an hour). This does not seem right - does anyone know what the problem might be or where I can look for a solution.

Thanks

[KCS_1]

hi,Arman.

It's may good thread for your LTO's performance , I think

Have a look at this


http://www.hp.com/products1/storage/products/tapebackup/ultrium_tapedrives/ultrium230/infolibrary/whitepapers/



Good luck!

[Michael Tully]

What method/software are you using now for the sybase backups?

Is this a directly connected tape drive to the N?

From performance reports you should be experiencing a throughput of 80Gb +/- an hour.

[Eugeny Brychkov]

Arman,
to allow LTo 230 working in the best way host should provide data streaming at speed at least 15MB/s. Your issue is most possible caused by little block size (2KB). It should be 32K the best
Eugeny

[Steven E. Protter]

Real life experience, same drive, L Class system.

We can back up 24 Gig with some files open in 20-25 minutes depending on how active the system is.

fbackup or Ignite.

These pups are fast and spec out around 80 Gig an hour.

Time to check your patching. December 2002 is the way to go minimally. You also want to run a patch search for stape and tape performance. I recalling having an issue that was resolved with one PHSS patch.

For 11.11 consider:
PHCO_23914 Enhancement support for Ultrium Tape
PHCO_24456 Shared tape administration
PHKL_28217 SCSI tape culmulative

I know you are 11.00 but there are probably equivalents for 11.00 for these patches. I don't have any 11.00 systems with Ultrium drives.

If you find the equivalents you might solve your problem. Here is my search on an 11.00 system but I know the Ultrium patch isn't there.

[1728]# swlist -l product | grep tape
PHCO_16720 1.0 st(1) New shared tape administration comm
and
PHKL_14026 1.0 mtstart() Page fault,tape0 Hang,Panic pat
ch
PHKL_22161 1.0 SCSI Tape (stape) cumulative
PHKL_22287 1.0 SCSI Tape (stape) cumulative
PHKL_22856 1.0 SCSI Tape (stape) cumulative
PHKL_23790 1.0 SCSI Tape (stape) cumulative
PHKL_24221 1.0 SCSI Tape (stape) cumulative
PHKL_26784 1.0 SCSI Tape (stape) cumulative
PHKL_28216 1.0 SCSI Tape (stape) cumulative

Hey its a start right? We can't do all the work.. :-)


SEP

[Arman Mahboobi]

Thanks,

I've tried changing the blocksize and backing up - this seems to have significantly improved the dump process, however, the load process improved to about 4 gb per hour.

We use sybase v11.5, we tried using sybase 12's backup and the load performance increased to about 15 gb an hour.

We will be looking at installing the latest patchs soon.

Thanks again,
Arman

[Rothery Harris]

Dear Arman,

Increase the blocksize to 256KB. It should be at least 64KB. You will see improvement up to about 64KB. Also look at your filesystem. There are some tuning methods available. To acieve creditable performance you must have a fast disk RAID.
Also check that the SCSI negotiation is set at 'unlimited' for the HBA. To do this reboot the system and enter the service menu from main.(This is the menu before ISL) There is a SCSI command which will set the speed. The syntax is something like SCSI> SPEED <> UNLIMITED

Rothery