ADSM-L

Re: Autochangers - device support and recommendations

1996-11-26 22:28:15
Subject: Re: Autochangers - device support and recommendations
From: Joe Morris <morris AT UNC DOT EDU>
Date: Tue, 26 Nov 1996 22:28:15 -0500
On Wed, 27 Nov 1996, Zvi Bar-Deroma wrote:

| I have to make a quick decision concerning the purchase of an
| autochanger of some-magnetic tapes for use with ADSM. The server...

| 1. To ADSM development - #4 (3570, the Magstar MP) is not on your h/w
|    list (on the WEB); as for the Quantum drives (#5,6), other models are
|    mentioned there (2000T, 2000XT, 4000). What are the plans for availability
|    of support for these Changers (mainly for #4 ) for AIX servers ? Please
|    specify dates (even if estimated/approximate).

I'll let them handle this, but I think you'll be OK.

| 2. Are there any user experiences with these drives (good or bad)  in
|    conjuction with ADSM usage, as well as any other issues.

Definitely go with DLT.  I've been using it for years.  Unless the job is
small, stay away from helical scan technology--it's reliable, but not
nearly as reliable or as fast as DLT.  I've been running a DLT 4700 with
ADSM flawlessly.  I'm considering the purchase of a drive from Overland
Data that seems to have some nice, scalable DLT libraries.

| 3. Can anyone shed light on the reliability of the 3 1/2 technologies
|    of the above drives (8mm. + 8mm. Mamoth, DLT and Magstar) ?

DLT is fast, reliable, and proven technology.

| 4. What (if at all) is the merit of 2 vs. 1 drive ? Can ADSM exploit
|    this feature where available (I guess the 2 issues can be involved -
|    speed and reliability) ?

You'll need to ask youself some questions like, can I afford to be without
a tape drive if one fails?  Do I need to consolidate volumen information
periodically (requires 2 drives)?  Do I need to replicate tapes?  What if
disaster strikes?  Your speed should increase as long as you setup the
drives architecturally (sp?) right.

| 5. Anyone willing to briefly explain (or refer me to some url) about
|    the differences between the "hellical scan tape" vs. Linear tapes ?

Oh boy, here goes.  If anyone has a URL, that would be great, but I don't
know of one at the moment.

Helical scan--The tape head writes diagnolly across the tape medium.  Here
is crude example in good ol' ASCII:

=================================
/ / / / / / / / / / / / / / / / /
 / / / / / / / / / / / / / / / /
/ / / / / / / / / / / / / / / / /
 / / / / / / / / / / / / / / / /
/ / / / / / / / / / / / / / / / /
=================================

As you can imagine, the spacing is very tight.  The head spins at a high
rate of speed and does create a fair amount of friction.  For performance
concerns, it's critical to keep data streaming to the tape drive.  If
there are gaps, the tape will stop, the head will pull away from the tape
to stop friction in one spot.  When the data starts flowing again, the
tape must be rewound a little bit, then spin-up the head, get the head
synchronized with the tracks on tape, and continue on.  This is a big
performance hit when this ocurrs frequently.  The head actually has at
least 2 heads--1 for wrting, and 1 for reading and they are offset from
each other.  Therefore, after a track has been written and the write head
starts to travel along the back side away from the tape, the read head is
getting ready to verify the just written track.

Linear technology--This is also referred to as serpentine formatting as
the data is written in tracks on tape end to end.  Many tracks can be
written.  Here is another crude example of DLT:

================================
------->--------->-------------+
                               |
                               |
+---------<---------<----------+
|
+------>--------->--------------
================================

There are two heads on DLT--1 for wrting, and 1 for reading, but they swap
roles as they are working.  The heads are in-line with each other.  The
head begins writing in one direction with the forward head writing and the
trailing head verifying the data in a single track.  When the end of the
tape is reached, the tape simply starts moving in the other direction and
write another track parallel to the one it just wrote.  Of course, the
heads reverse their roles in the read/write process.  When data flow
stops, the tape stops immediately and only need to backup just a little
bit.  When data starts to flow again, it takes off very rapidly as it does
not have to go through the synchronization process Helical does.  Also
bear in mind that files can be found much faster in restoring files as DLT
can simply "hop" to the appropriate track to get at the file faster.

Hope that helps...
_______________________________________________________________
Joe Morris  -  morris AT unc DOT edu  -  http://sunsite.unc.edu/morris
Academic Technology and Networks (formerly OIT), Development
University of North Carolina at Chapel Hill