Everhart,Glenn
From:	Everhart,Glenn
Sent:	Tuesday, April 28, 1998 8:18 AM
To:	Info-VAX@Mvb.Saic.Com
Subject:	RE: "Fast-forward" on DLT drives
The reason fast forward isn't so fast prior to vms 7.1 is that
VMS used skip by record, which gives basically similar speed,
except for DLT, to skip by file, yet allows the skip file
semantics to be preserved. Attempts to use skip by filemark
have been tried since V5.5 and hit problems. One of the more
knotty ones is that ANSI format means that double EOF does not
mean EOT, and third party packages that write info beyond
double EOF on ANSI tapes get screwed up by skip by filemark.

Another knotty problem is the fact that if you have 8000 files
(for example) on a SCSI tape drive, and tell it to skip 8000 file
marks, you get a success, not a blank tape error. Yet you are
required to leave the tape positioned between the final double EOF
marks after a skip. This means that you have to skip backwards,
check for EOF mark, skip back again, check again, and skip
forward appropriately after each file positioning. This can be
slower than the skip-by-record strategy and whether it is basically
cannot be determined ahead of time without knowing what's on the
tape. Sue Sommer finally worked this out, in all its gory detail, for
V7.1, but the problem is much harder than it appears at first blush.


> -----Original Message-----
> From:	Terry Kennedy [SMTP:terry@spcunb.spc.edu]
> Sent:	Tuesday, April 28, 1998 7:23 AM
> To:	Info-VAX@Mvb.Saic.Com
> Subject:	Re: "Fast-forward" on DLT drives
> 
> Scott Vieth <svieth@ameritech.net> writes:
> > We've got our DLT drives connected to KZMSA ports on our DEC
> 7000s.This should
> > work, right?
> 
>   I bailed out of DEC stuff before 7000's, so I'm not sure. In
> general, the
> busless workstation-style boxes had direct SCSI controlled by the host
> driver,
> while systems with SCSI cards attached to the system bus spoke TMSCP
> to the
> host and SCSI to the drive (the former using the MKDRIVER, the latter
> the 
> MUDRIVER).
> 
> > Why do the StorageWorks folks keep crowing about the "fast forward"
> feature if no
> > one can
> > get it to work?  I'm puzzled.  Seems to me like they should "put a
> cork in it"
> > until they get the darn thing to work.  I'm betting that I'm not the
> only one
> > who had left a StorageWorks seminar with high hopes and then found
> > that it still takes 15 minutes to position a tape when trying to
> restore a file (on
> > DLT).
> 
>   Beats me, except that it's common for hardware features to exist
> long before
> there is operating system code to take advantage of them.
> 
> 	Terry Kennedy		  Operations Manager, Academic Computing
> 	terry@spcvxa.spc.edu	  St. Peter's College, Jersey City, NJ
> USA
>         +1 201 915 9381 (voice)   +1 201 435-3662 (FAX)