🚀 go-pugleaf

My site just recently purchased a Quantum DLT4700 mini-library from
Applied Digital Systems (ADS).  The drive is connected via a Fast SCSI-2
Narrow/Differential SBus adapter to a Sun SPARC-10 running SunOS 4.1.3.

I have two (well, three) concerns that I'm hoping someone can help me
with: 1) Hardware documentation, 2) Proper system definition, and 3) I/O
Performance.

1) Hardware documentation.  Other than a vibrant-pink "READ ME FIRST"
   pamphlet and a fold 'em up quick reference card, the only documentation
   that arrived with the tape drive was a spiral-bound photocopy of DEC's
   "DLT4000/DLT4500/DLT4700 Series Magnetic Tape: Product Manual" (order
   number EK-TH5XX-IM).  This document is a Preliminary Release, dated
   June 1994!  My question for other recent DLT4700 purchasers (we ordered
   ours on August 4th, it arrived September 11th): Did you get the same
   pitiful hardware documentation with your drive, as I did???  If not,
   and you received a more recent version of the Product Manual or something
   entirely different, then please e-mail me with the manual(s) order number(s).

2) System definition.  I added the following definition to the 
   /usr/sys/scsi/targets/st_conf.c file on my system:

   /* Quantum 1/2" Cartridge */
   {
           "Quantum DLT4700", 15, "Quantum DLT4700", ST_TYPE_HIC, 16384,
           (ST_VARIABLE|ST_BSF|ST_BSR),
           400, 400,
           { 0x80, 0x81, 0x82, 0x83 },
           { 0, 0, 0, 0 }
   }

   This information was obtained from the vendor (ADS) via e-mail.  The
   only thing I changed was "DLT4000" to "DLT4700", as that is the product
   id being returned in the INQUIRY data by the drive (according to a
   probe-scsi-all that I did from the boot PROM).  This information is
   also pretty close to that provided on the Quantum WWW site (I believe
   the only difference is that I've listed the density codes in low-to-high
   order, where Quantum lists them in high-to-low order).

   After adding the definition I rebuilt the SunOS kernel, rebooted the
   machine and noted that the kernel recognized and reported the presence
   of the DLT4700 (at /dev/rst5: it is SCSI ID 5 on the second SCSI bus).
   I am able to talk to the drive using device names /dev/rst5, /dev/rst13,
   /dev/rst21, and /dev/rst29.  When I use rst5 or rst21, the front-panel
   20.0 GB LED is the only density indicator that lights up.  When I use
   rst13 or rst29, the Compress LED lights up in addition to the 20.0 GB LED.
   I'm using CompacTape IV cartridges.  This is not quite what I expected.
   I expected rst5 to light up the 10.0 GB LED, rst13 to light up the 10.0 GB
   and Compress LEDs, rst21 to light up the 20.0 GB LED, and rst29 to light
   up the 20.0 GB LED and Compress LED.  Will I only get the 10.0 GB LED to
   light up when I'm using CompacTape III cartridges?  Does the drive know
   I'm using CompacTape IVs and will always set itself to 20 GB mode?

   Anyway, the other question about the system definition I have is why do
   both Quantum and the vendor from which I purchased the drive indicate
   that the block size of the device should be set to 16 KB (i.e. 16384)?
   The drive is supposedly capable of handling block sizes up to 16 MEGAbytes.
   Has anyone else tried bumping up the block size to something larger,
   e.g. 64 KB (65536)?  Or does it even matter?  That is, if I specify a
   block size of 63 KB (the "b" flag on the "dump" command with a value of
   126) when I do my dumps, will that override the 16 KB value specified in
   the system definition?

3) I/O Performance.  I did some trial level 0 dumps of several filesystems
   using the SunOS /etc/dump command.  The format of the command I used was:

   /etc/dump 0bdsf 126 62500 28000 /dev/nrst29 /

   and etc.  Note that I'm specifying the density as 62500 bpi, and the
   tape "size" as 28,000 feet.  (Sure would love to know what values other
   people are using.)  The blocking factor I'm using is 126 512-byte blocks
   (i.e. 63 KB).

   The transfer rate I observed for the dumps were rather abysmal: varying
   from 0.27 MB/sec to 0.54 MB/sec!  Note that four of the filesystems I
   dumped are located on disks that share the same SCSI bus as the DLT4700:
   that didn't seem to make much difference however.  One of those filesystems
   showed the best transfer rate (0.54 MB/s).

   Admittedly, none of the filesystems I'm dumping are either very large
   or very full (the largest dump only wrote 895.57 MB to tape).  I would
   expect larger dumps to get a bit better transfer rates, but even so I
   still doubt that they would get much closer to the 1.25 MB/s *native*
   (uncompressed) transfer rate that Quantum claims for the DLT4x00.

   By the way, the "scsiinfo" utility indicates that the tape drive is
   capable of 10.0 MB/s synchronous transfer:

   esp1: st5 tgt 5 lun 0:
           Synchronous(10.0MB/sec) Clean CanReconnect
           <<< SCSI probe failed: device driver doesn't support probing >>>

   So, does anyone have any clues?  Is Quantum shipping DLT4x00s from the
   factory with Buffered Mode disabled?  Is my drive waiting for each block
   to be successfully written to tape before sending a completion message
   back to the operating system?

   I hope someone can help...we were looking forward to the arrival of
   our first DLT tape drive and are somewhat disappointed that we're only
   getting Exabyte 8500 transfer rates out of it!  :-(
 

Thanks for any help you can provide,

 - Lee

P.S.  Can anyone tell me where I can grab a copy of Quantum's "DLT Tools"
      package?  I couldn't find it on their Web site.

+--------------------------------------------+--------------------------------+
| Lee Garner (Sr. Systems Programmer)        | Internet:  garner@acns.fsu.edu |
| Academic Computing & Network Services      | BITNet:    Not any more!       |
| Florida State University                   | Facsimile: (904) 644-8722      |
| Tallahassee, Florida 32306-3042  USA       | Voice:     (904) 644-2591      |
+--------------------------------------------+--------------------------------+