A quick guess is that you are using the 12H Autoraid without the ARM utilities
in performance mode. The 12H is slow due to raiding the drives. The 12H gives
you protection against disk failure but you pay on performance.

Store to disk with autoraid as you have noticed is not going to improve unless
you put the drives in "performance mode" in a non-protected array.
My advise would have been to mirror disks vs raid.
YGWYPF "you get what you pay for"

Sorry
> I am looking for some suggestions why it would be slower to write to a
> non-system volume set (5 SCSI discs) running on the MFIO SCSI channel
> (external) versus a DAT3 on exactly the same SCSI channel (internal).
>
> 928RX
>
> The system volume set is running on an add-on FWD SCSI 2 card. AutoRAID 12H.
> Nothing on the system volume set is running over the MFIO card SCSI channel.
>
> This is for backup using Orbit Backup/iX.
>
> On the MFIO card SCSI channel there is both the system tape drive - a DDS3
> drive (internal ribbon cable), and a set of five SCSI discs in a non-system
> volume set "USER_VOL_SET_1" (external SCSI cable).
>
> When Orbit Backup/iX backs up the entire system volume set onto DAT tape on
> the DDS3 drive, it takes about 2 hours.
>
> When exactly the same backup goes to a disc file "FULL.BKUP.SYS" on the
> non-system volume USER_VOL_SET_1, it takes about 3.5 hours. 75% longer.
>
> The only group on the non-system volume set "USER_VOL_SET_1" is "BKUP.SYS".
> The backup file set specifically excludes this group ([log in to unmask]) plus the
> store command has "ONVS=MPEXL_SYSTEM_VOLUME_SET;". Which should prevent
> anything from USER_VOL_SET_1 being part of the backup.
>
> The way I see it, for file reads (the files the backup is storing), plus any
> control files written to, the amount of IO should be exactly the same going
> in/out of the system volume set (AutoRAID 12H on separate FWD SCSI card).
>
> The output to the store file(s) in terms of number of IO's on the MFIO SCSI
> channel should be the same whether it is going to a tape drive or a set of
> five disc drives. Shouldn't it? Or am I missing something here?
>
> I know the DDS3 has hardware compression. So the actual number of bits
> written to tape would be less than to disc. However, this compression
> happens inside the DAT3 drive. So the same amount of information should be
> flowing down the SCSI connection shouldn't it?
>
> Even though a DDS3 is a pretty dang efficient tape device, I would still
> think even older 2GB and 4GB SCSI drives would be faster. Is my thinking
> wrong on this?
>
> So any ideas what would cause this huge difference? Allocation of disc
> space/extents? Buffering? Other?
>
> Any ideas how to make the backup to disc as efficient as the backup to tape?
>
> * To join/leave the list, search archives, change list settings, *
> * etc., please visit http://raven.utc.edu/archives/hp3000-l.html *

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *