burow@physics.utoronto.ca (Burkhard Burow) (02/12/90)
Digital Audio Tape (DAT) for Data Storage
-----------------------------------------
Burkhard Burow, Dept. of Physics, Univ. of Toronto.
burow@oldkat.physics.utoronto.ca
[usual disclaimers. Please e-mail corrections and/or additions to this note
to me, and I will repost.]
I've condensed most of the following info. from these articles:
DAT for Data Storage, IEEE Spectrum, October 1989, pp.34-38.
DAT's The Future of Tape, Unix World, January 1990, p.184.
Peripherals, IEEE Spectrum, February 1990, p.30.
The Data/DAT system, and the reasoning for it is described in
A New Concept of Data/DAT System, IEEE Transactions on Consumer Electronics,
v. 35, n. 3, pp. 660-671.
My research group currently uses Exabyte Drives, so I'll compare their
capabilities to those of the DAT drives.
DAT Drive Exabyte Drive
--------- -------------
Hardware:
---------
Tape: DAT audio 8mm video
(Comment: Their size and cost difference is negligible.)
Performance:
------------
Storage capacity per tape. [Gbytes]
1.3 2.3
(Comment: This will be for the DDS system, the Data/DDS has ~20% less space.)
Data transfer rate, i.e. Reading from / writing to continuous segment
of the tape. [Kbytes/second]
183 246
Access time, i.e. time to locate arbitrary position on the tape.
avg.: 20 seconds 1-10 minutes
max.: 40 seconds 1-10 minutes
(Note: Exabytes are fantastic backup systems, but the access time hampers
their use for 'data'.)
Data Reliability:
-----------------
Error rate: <10E-15 ???
(Note: The DAT's error rate seems to be worse if one writes hundreds of
times onto the same spot inside the tape. The Data/DAT
format is supposed to overcome this problem, but there are
probably very few applications where one would overwrite
the same segment of the tape this often. Therefore for
the vast majority of applications, the DDS format is extremely
robust, but reader's can consult the above Data/DAT paper for
more information.)
Standards:
----------
Exabyte is a product from a single company and hence there is a single
data format at the byte level. 50 000 Exabyte drives are currently in use.
Question: What is the compatibility between Exabyte tapes read and written
on VAXen and unix boxes. Is it similar to the reel-to-reel story?
DAT has two major formats. DDS and Data/DAT. DDS, by HP and Sony with
licences to 13 other vendors, has a years head start,
is faster and holds more data per tape. Data/DAT, by Hitashi and Toshiba
claims that DDS doesn't have enough error correction, (see Errors: above),
and replaces about 16% of DDS data capacity with more correction bytes.
There also exist a subset of Data/DAT which provides for random write
access to the tape. This format sacrifices a total of 28% of the data
capacity on a DDS tape.
Cost:
-----
Drive, controller, and software:
US$6000-$8000 Cdn$5000
(subtract for MSLP (approx. price paid as part
exageration and of a package deal)
edu. discount)
After looking at the numbers, it looks like the only reason to use
DAT's, with current performance standards, instead of Exabytes is
the difference in access time. The access time for the Exabyte could be
probably be overcome if one buffered the tape with a ?00 Mbyte disk,
a cost effective solution until DAT's come down in price or improve in
performance.
Drivers are an issue not addressed in the above discussion. It doesn't
matter for system backups, but Exabyte drivers are apparently very good
at tying up unix machines. With so many manufacturers in the DAT race
I'd imagine their drivers to be good. Drivers are not an issue for VAXen
since the SCSI cards on the Q-bus, etc. should provide proper buffering.
Please point out errors to me, and I will repost.