I was asked by Richard Rhodes to describe reliability of the various tape
technologies. Before I knew it, it turned into a dissertation on the
subject:
I'll describe data reliability and then drive reliability.
Most reliable: 3590. The technology makes this so. The tapes use an
internal servo track to ensure proper head alignment on read and write
operations. This is somewhat unusual for tape, but is similar to what has
been used in disk technology for some time. Tape never touches anything
except tape while in the drive itself. The tape rides an air bearing
through the drive. Similar to DLT, the tape end is removed from the
cartridge and "taken" to the head, but on it's way, it never touches
capstans or any other piece of the drive. Way cool, but as you can imagine,
hard and expensive to do. This contributes to the cost of the drive. Due
to the embedded servo, 3590 cartridges can not be deguassed as an erasure
strategy. I don't have the numbers, but I would think the stated
reliability and readability for the media is somewhere in the 50K passes and
25-50 year life.
Next: DLT. No embedded servo. The heads are positioned using the data
track itself. A little less precise than with a servo, but probably changes
the bit error rate only slightly, perhaps from 1 in 10 to the 23 to 1 in 10
to the 22 or something. Still a way small number. Tape touches lots of
stuff as it is passed from the cartridge to the head so there is more tape
wear in DLT than in 3590 so tape life isn't as long. Probably still rated
as high (check the actual specs on the Quantum site), but in actuality, I'd
be a bit more skeptical. However, how many times is a tape actually written
50K times (I know, I know, with serial serpentine, the number of passes over
the tape is much higher than the apparent number of passes, but still a long
way from 50K for most tapes).
Personal Experience: I've seen a number of DLT tapes with leader problems.
I've also seen a number of DLT drives with the leader fishing mechanism
(that's a highly technical term) broken. Not hard to fix, but the drive is
down while you have it fixed. I've also experienced problems with tapes
that have been dropped. I do not have the same experience with 3590, but
I'm sure we'll get some experiences like that now.
AIT: This is the new kid on the block. The technology is very similar to
8mm. That is, the tape is wrapped around a rapidly spinning head. Can you
say belt sander? In the case of 8mm, this was a major problem and
contributed to short tape and drive life and the constant worry about
whether any data written would be read. (4mm uses the same basic idea, but
the wrap angle, the amount of tape actually touching the head, was less, so
the amount of wear was less). AIT overcomes this by using Advanced metal
evaporative (AME) tape. Rather than painting the magnets on the plastic
substrate, the magnets are deposited on the plastic in a highly clean vacuum
chamber. The tape wears much less, if at all. Think of a piece of chromed
metal and you have the idea. I've lost a great deal of my skepticism about
this technology as I've used it and studied the literature. The URLs below
point to some interesting reading:
http://www.qualstar.com/146100.htm
http://www.qualstar.com/146014.htm
Drive reliability seems good as there are few moving parts compared to DLT
or 3590.
There are new technologies on the horizon as well.
LTO: Linear Tape-Open. Developed by IBM, HP and Seagate. More info in the
links below.
IBM Ultrium: Similar to 3590, but using LTO guts.
http://ssdweb01.storage.ibm.com/hardsoft/tape/lto/prod_data/ultrium.html
http:\\www.lto-technology.com
IBM Accelis: Similar to 3570, but using LTO guts.
http:\\www.lto-technology.com
Qualstar showed products at Comdex based on the Ultrium drives. I have no
idea about the cost, but I will post something when I do.
There, that should keep everyone reading and studying for a good long while.
Please note: As always, these are my opinions based on my own observations
and experiences. They do not represent the views of SSSI.
Kelly J. Lipp
Storage Solutions Specialists, Inc.
PO Box 51313
Colorado Springs CO 80949
719-531-5926
www.storsol.com
lipp AT storsol DOT com
|
