[comp.sys.mips] Experience with DAT drives?

k2@charly.bl.physik.tu-muenchen.de (Klaus Steinberger) (03/06/90)

Dear Netlanders,

has anybody tried to use a HP/Sony DAT Drive or a Wangtek drive
on a MIPS System?
We are thinking about buying such things instead of Exabyte.

Sincerely,
Klaus Steinberger

Klaus Steinberger               Beschleunigerlabor der TU und LMU Muenchen
Phone: (+49 89)3209 4287        Hochschulgelaende, D-8046 Garching, West Germany
BITNET:  K2@DGABLG5P            DFN:     k2@t30.physik.tu-muenchen.dbp.de
Bayerisches Hochschulnetz: k2@charly.bl.physik.tu-muenchen.de

roe@sobeco.com (r.peterson) (03/12/90)

From article <1273@tuminfo1.lan.informatik.tu-muenchen.dbp.de>, by k2@charly.bl.physik.tu-muenchen.de (Klaus Steinberger):
> Dear Netlanders,
> 
> has anybody tried to use a HP/Sony DAT Drive or a Wangtek drive
> on a MIPS System?
> We are thinking about buying such things instead of Exabyte.
> 

I've used the Wangtek drive on an M-120 with no problems at all.  Just make
sure to set the jumpers on the DAT for the same disconnect size as a
builtin 150MB Archive drive.

However, both the transfer rate and storage capacity are about half the
exabyte.
-- 
"Norf gribble mitz der bork bork bork?"        Roe Peterson
"Lick spittle dwing dtz fritz slppp!"          roe@sobeco.com
	- Uncle Duke                           uunet!sobeco!roe

yar@cs.su.oz (Ray Loyzaga) (03/14/90)

In article <1990Mar11.160020.27837@sobeco.com> roe@sobeco.com (r.peterson) writes:
> From article <1273@tuminfo1.lan.informatik.tu-muenchen.dbp.de>, by k2@charly.bl.physik.tu-muenchen.de (Klaus Steinberger):
> > has anybody tried to use a HP/Sony DAT Drive or a Wangtek drive
> > on a MIPS System?
> 
> However, both the transfer rate and storage capacity are about half the
> exabyte.
I have used a JVC DAT called a Gigatape, it plugs straight into the
SCSI port and works well on the M/120, there seems to be a firmware
mismatch once it is on the system that makes the M120 fail the cold
boot diagnostics, this doesn't stop the device from being usable, but
should be fixed by mips and the tape manufacturer ...
The nice thing about this drive is that it has a display that gives
constant info about what is going on, seeking, writing etc, with
a number that increments as the tape is used, so you can get perfectly
accurate numbers about how much of the tape has been written.
The speed of writing was slower than the Exabyte, but certainly not
by half, maybe 80% as fast, the real win is that the tape can do end to end
seeks in 10-20 seconds, whereas the Exabyte is absolutely painful
with each command. The Gigatype, being fully digital, will be more
reliable and it does not lose sync due to a tape error e.g.
If there is a media error early in the tape, the drive will still
be able to find any later files without trouble. This is great for
multi-machine backups and restores.

In my opinion, the DAT style systems are much better to use for
multi-file tape use, most of this comes from the speed of positioning.
The major drawback being the requirement for pre-formatting the tape,
which can take >3hours each. Maybe they will sell preformatted tapes?
Or maybe they will get ones that format on the fly.

roe@sobeco.com (r.peterson) (03/28/90)

From article <766@cluster.cs.su.oz>, by yar@cs.su.oz (Ray Loyzaga):
> In article <1990Mar11.160020.27837@sobeco.com> roe@sobeco.com (r.peterson) writes:
>> 
>> However, both the transfer rate and storage capacity are about half the
>> exabyte.

> I have used a JVC DAT called a Gigatape, it plugs straight into the
> SCSI port and works well on the M/120, there seems to be a firmware
> mismatch once it is on the system that makes the M120 fail the cold
> boot diagnostics, this doesn't stop the device from being usable, but
> should be fixed by mips and the tape manufacturer ...

(This is interesting to start with, since Gigatape is already a
 tradename for OEM exabyte drives...)

Hmmm... I've had no problems whatever with the WANGDAT drive - passes
all boot tests, and can be used to both boot and install the
operating system (both miniroot and the full distribution).

> The nice thing about this drive is that it has a display that gives
> constant info about what is going on, seeking, writing etc, with
> a number that increments as the tape is used, so you can get perfectly
> accurate numbers about how much of the tape has been written.

Ahah!  A blinkenlights man!  After my own heart, but:

Why is this information useful?  The same thing is available on several
OEM versions of the exabyte drive, but I find it meaningless.  Either
a write works or not.

What do you plan to do?  Hire someone to watch the front panel of the
tape drive and hammer the control-c if the drive is nearly full?

Furthermore, (at least on the exabyte), all such statistics are available
with SCSI inquiry commands.  Much better to grab error stats and log
them than depend on users noticing the front panel display.

[ What follows is an unabashed plug for exabyte drives. ]
[ I do not work for exabyte, but [standard disclaimer ]].

> The speed of writing was slower than the Exabyte, but certainly not
> by half, maybe 80% as fast.

True, given that you are using the standard mips scsi device driver.
The exabyte drive transfer speed is a real big win in buffered mode;
simply modify the (mips-supplied) device driver to set the
buffered mode bit, and you will see *real* drive transfer rates.
We get between 242KBS and 244KBS (the drive itself is rated at
246).

DAT just simply, physically can't read/write faster than about
180KBS. (Specs say 176 KBS for the WANGDAT).

> the real win is that the tape can do end to end
> seeks in 10-20 seconds, whereas the Exabyte is absolutely painful
> with each command.

Why is this a real win?  How many backup/restore utilities take
advantage of seek time?  Who cares how fast a drive rewinds?
Also, I should point out that the DAT drives can *not* perform
block seeks from end to end in 20 seconds - they can search
file marks this fast.

We've set up our system to use exabytes in 100MB savesets;  worst-
case restore time (the last file on the last saveset) is about
18 minutes (forward space 19 files, read the balance, and grab the
last file).

> The Gigatype, being fully digital, will be more
> reliable and it does not lose sync due to a tape error.

This is incorrect.  The error correction on an exabyte drive provides
extensive ECC, with read-after-write correction (with no loss whatever
in terms of write speed).  Simply being "fully digital" has nothing
whatever to do with reliability.

Also, any tape drive (digital or not) can "lose sync" if the tape is
garbled enough (magnetic field, physical damage, etc).  The exabyte simply
does not "lose sync" without major tape damage.

> If there is a media error early in the tape, the drive will still
> be able to find any later files without trouble. This is great for
> multi-machine backups and restores.

Well, I've a problem with this statement too.  Having used exabytes
extensively for the last 2+ years, without *one* read error, your
point simply isn't.

Also, an exabyte drive is probably your best bet for recovery of
bad tape spots.  Check out the extensive error detection/recovery that
is automatically encoded by the drive.

Furthermore, in worst case, most dump/restore systems will simply
choke and die when a bad block is encountered.

> In my opinion, the DAT style systems are much better to use for
> multi-file tape use, most of this comes from the speed of positioning.

Again, my question remains - what dump/restore utility uses seek?
What tape device driver (ATT 3B2 aside) supports it?
Also, if an exabyte can forward space the whole tape in less than
15 minutes, who cares?

-- 
"Norf gribble mitz der bork bork bork?"        Roe Peterson
"Lick spittle dwing dtz fritz slppp!"          roe@sobeco.com
	- Uncle Duke                           uunet!sobeco!roe

k2@charly.bl.physik.tu-muenchen.de (Klaus Steinberger) (03/28/90)

roe@sobeco.com (r.peterson) writes:

>From article <766@cluster.cs.su.oz>, by yar@cs.su.oz (Ray Loyzaga):
>> In article <1990Mar11.160020.27837@sobeco.com> roe@sobeco.com (r.peterson) writes:
>> the real win is that the tape can do end to end
>> seeks in 10-20 seconds, whereas the Exabyte is absolutely painful
>> with each command.

>Why is this a real win?  How many backup/restore utilities take
>advantage of seek time?  Who cares how fast a drive rewinds?
>Also, I should point out that the DAT drives can *not* perform
>block seeks from end to end in 20 seconds - they can search
>file marks this fast.

>We've set up our system to use exabytes in 100MB savesets;  worst-
>case restore time (the last file on the last saveset) is about
>18 minutes (forward space 19 files, read the balance, and grab the
>last file).

You are thinking that there is nothing that can be done with a
DAT or Exabyte besides the standard backup/restore tasks!

We want to use DAT's for Data-acquistion, and there is very much
pain if you need a long time for seeking your files. Also there
is very much pain if you need 15 Minutes to rewind your tape!

Also there is only one manufacturer for EXABYTE, namely EXABYTE.
The DAT's you can get from many suppliers. This is very important,
if you depend on the availability, say in 5 or 10 years!

Sincerely,
Klaus Steinberger
Klaus Steinberger               Beschleunigerlabor der TU und LMU Muenchen
Phone: (+49 89)3209 4287        Hochschulgelaende, D-8046 Garching, West Germany
BITNET:  K2@DGABLG5P            Internet: k2@charly.bl.physik.tu-muenchen.de

chris@bclwn.bcl.co.nz (Chris Mackerell) (03/30/90)

In article <1990Mar28.053504.14178@sobeco.com> roe@sobeco.com (r.peterson) writes:
>
>
>True, given that you are using the standard mips scsi device driver.
>The exabyte drive transfer speed is a real big win in buffered mode;
>simply modify the (mips-supplied) device driver to set the
>buffered mode bit, and you will see *real* drive transfer rates.
>We get between 242KBS and 244KBS (the drive itself is rated at
>246).


How?  Information please!!

>
>Also, an exabyte drive is probably your best bet for recovery of
>bad tape spots.  Check out the extensive error detection/recovery that
>is automatically encoded by the drive.
>

Unless, like me, you got a box of dud tapes from Sony!



I'm a fan of Exabyte units (we sell both 8mm & DAT units).

We are finding that DAT hasn't the capacity for some of our customers -
they are complaining about the Exabytes having 'only 2.3GB'...


Has anyone got the new 5GB Exabyte ? Any comments on it? (I've only
read about it in the press).


Cheers,

Chris.

BBBB   CCC  L     Business Computers Ltd               email: chris@bcl.co.nz
B   B C     L     P.O. Box 3236                        phone: + 64 4 801 6716
BBBB  C     L     Wellington 6000                      fax:   + 64 4 852 752
B   B C     L     New Zealand                          Chris Mackerell
BBBB   CCC  LLLL  Also at Auckland & Christchurch      Support Manager  

      UNIX Support - It's not just a job, it's an adventure!
(Unless stated otherwise opinions expressed are my own, not those of BCL)




-- 
BBBB   CCC  L     Business Computers Ltd               email: chris@bcl.co.nz
B   B C     L     P.O. Box 3236                        phone: + 64 4 801 6716
BBBB  C     L     Wellington 6000                      fax:   + 64 4 852 752
B   B C     L     New Zealand                          Chris Mackerell

yar@cs.su.oz (Ray Loyzaga) (03/30/90)

In article <1990Mar28.053504.14178@sobeco.com> roe@sobeco.com (r.peterson) writes:
> Hmmm... I've had no problems whatever with the WANGDAT drive - passes
> all boot tests, and can be used to both boot and install the
> operating system (both miniroot and the full distribution).
> 

The drive I tested made the SCS master test fail.

> Ahah!  A blinkenlights man!  After my own heart, but:
> 
> Why is this information useful?  The same thing is available on several
> OEM versions of the exabyte drive, but I find it meaningless.  Either
> a write works or not.

We have type 8mm video tape backup devices, neither supports anything
other than a red and a green light, the green light turns on
when the tape is correctly inserted, the red light tells you there
is power connected. I find the Gigatape display more useful because
I like knowing what the tape is doing, like is someone writing it?
is it rewinding, has it stopped, specially since the drive locks up
sometimes requiring a system reboot.

> What do you plan to do?  Hire someone to watch the front panel of the
> tape drive and hammer the control-c if the drive is nearly full?

Very funny, I was trying to show that this info is very useful because
it allows you to get the numbers right for systems dumps, you know exactly
how much can be written, we got these numbers from the exabyte
by using lots of dd's with different block sizes and waiting till the tape
runs out, then backing off 10%. 
I find it useful to have information available that can tell you what is
happening at a glance, the "Exabyte" style drive we have does not give
enough information and the Gigatape gives more than enough, I know which
situation I prefer.
> Furthermore, (at least on the exabyte), all such statistics are available
> with SCSI inquiry commands.  Much better to grab error stats and log
> them than depend on users noticing the front panel display.
> 

I prefer instant feedback, we differ, I also run my terminal with
echo turned on.

> > the real win is that the tape can do end to end
> > seeks in 10-20 seconds, whereas the Exabyte is absolutely painful
> > with each command.
> Why is this a real win?  How many backup/restore utilities take
> advantage of seek time?  Who cares how fast a drive rewinds?
> Also, I should point out that the DAT drives can *not* perform
> block seeks from end to end in 20 seconds - they can search
> file marks this fast.

I care, I happen to search through tapes made up of lots of files
and I much prefer a system that allows me to stick the tape in
step many files to the end and take a look at the data contained
in a matter of seconds rather than the 8mm video system we have
that takes an inordinate amount of time to load, retension, seek.
I tried using it for an archival system that would search through
the tape and it took so long that I gave up, it was unusable, because
the file mark search time was so slow.

> We've set up our system to use exabytes in 100MB savesets;  worst-
> case restore time (the last file on the last saveset) is about
> 18 minutes (forward space 19 files, read the balance, and grab the
> last file).

Actually you are saying it takes 18 minutes to get to the last tape
file so that you can search it to see if it is on that tape,
if it isn't you need to load another tape which will take several minutes
to rewind and unload, then you have to load a new tape.
We cannot afford to keep a table of contents of the dumps on line, 
so we have to make some guesses based on the last change time of the file
to be restored.

> This is incorrect.  The error correction on an exabyte drive provides
> extensive ECC, with read-after-write correction (with no loss whatever
> in terms of write speed).  Simply being "fully digital" has nothing
> whatever to do with reliability.
> 
> Also, any tape drive (digital or not) can "lose sync" if the tape is
> garbled enough (magnetic field, physical damage, etc).  The exabyte simply
> does not "lose sync" without major tape damage.

I have lost big with these tapes before, many times, all info on later
tape files was also lost. The DAT tape works for seeks to later files I
know, I scratched the early part of a tape to test it.

shj@ultra.com (Steve Jay) (03/30/90)

chris@bclwn.bcl.co.nz (Chris Mackerell) writes:

>Has anyone got the new 5GB Exabyte ? Any comments on it? (I've only
>read about it in the press).

I called Exabyte & asked.  The 5GB drives won't be available until
summer or fall of this year.

Steve Jay
shj@ultra.com  ...ames!ultra!shj
Ultra Network Technologies / 101 Dagget Drive / San Jose, CA 95134 / USA
(408) 922-0100 x130	"Home of the 1 Gigabit/Second network"

roe@sobeco.com (r.peterson) (04/10/90)

From article <798@cluster.cs.su.oz>, by yar@cs.su.oz (Ray Loyzaga):
> In article <1990Mar28.053504.14178@sobeco.com> roe@sobeco.com (r.peterson) writes:
>> Hmmm... I've had no problems whatever with the WANGDAT drive - passes
>> all boot tests, and can be used to both boot and install the
>> operating system (both miniroot and the full distribution).
>> 
> 
> The drive I tested made the SCS master test fail.
> 

Hmmm - thats' kind of interesting - what drive and revision level are you
using?  Also, what boot proms?

> 
> We have type 8mm video tape backup devices, neither supports anything
> other than a red and a green light, the green light turns on
> when the tape is correctly inserted, the red light tells you there
> is power connected. I find the Gigatape display more useful because
> I like knowing what the tape is doing, like is someone writing it?
> is it rewinding, has it stopped, specially since the drive locks up
> sometimes requiring a system reboot.
> 

Is the drive itself locked?  What happens if you just reset the tape
drive (this can be a bit difficult if you have the exabyte mounted
in a cabinet with other devices)?  There is a possibility that the
device driver is the culprit.

>> What do you plan to do?  Hire someone to watch the front panel of the
>> tape drive and hammer the control-c if the drive is nearly full?
> 
> Very funny

Was meant to be, actually - I seem to have forgotten the :-).

> I was trying to show that this info is very useful because
> it allows you to get the numbers right for systems dumps, you know exactly
> how much can be written, we got these numbers from the exabyte
> by using lots of dd's with different block sizes and waiting till the tape
> runs out, then backing off 10%. 

That sounds like a safe way to estimate tape sizes, and should work fine
unless you get a very bad tape with a lot of invalidated blocks.
But, you don't _need_ a front panel display for this - dd will tell you
how much data was written.

>> Furthermore, (at least on the exabyte), all such statistics are available
>> with SCSI inquiry commands.  Much better to grab error stats and log
>> them than depend on users noticing the front panel display.
>> 
> 
> I prefer instant feedback, we differ, I also run my terminal with
> echo turned on.
> 

Personally, I prefer 5-bit with half duplex... :-).

The reason I prefer log files for this is that (at least at our site),
we usually don't have someone watching the tape drive during the backup
(or at any other time).  Log files permit later perusal of problem
conditions.

For instant feedback, just log stuff to the system console.

>> Who cares how fast a drive rewinds?
> 
> I care, I happen to search through tapes made up of lots of files
> and I much prefer a system that allows me to stick the tape in
> step many files to the end and take a look at the data contained
> in a matter of seconds [...]

I stand corrected here.  It is undoubtedly true that if you need to
search for files, the DAT drive will perform better.

> 
> Actually you are saying it takes 18 minutes to get to the last tape
> file so that you can search it to see if it is on that tape,
> if it isn't you need to load another tape which will take several minutes
> to rewind and unload, then you have to load a new tape.
> We cannot afford to keep a table of contents of the dumps on line, 
> so we have to make some guesses based on the last change time of the file
> to be restored.

I think the moral of the story here is that our needs and resources are
different.  We're backing up 4+ GB every friday, and we keep full logfiles
online at all times, so we don't need to guess where a file is located.
Our application requires high-speed read/write with high density, whereas
yours seems to require fast access to saveset "n" on the tape.

Being able to restore any file from a 2.3 GB tape within 20 minutes is
very acceptable at this site.

> I have lost big with these tapes before, many times, all info on later
> tape files was also lost. The DAT tape works for seeks to later files I
> know, I scratched the early part of a tape to test it.

I guess experience differs here too.  We've been using exabytes extensively
for quite a while without a read failure of any sort.  The only tape problems
we experienced was with a bad box of TDK tapes, and these were usable -
but could not record as much data (error correction kicked in a lot).  Sony
tapes have never given us a problem.
-- 
"Norf gribble mitz der bork bork bork?"        Roe Peterson
"Lick spittle dwing dtz fritz slppp!"          roe@sobeco.com
	- Uncle Duke                           uunet!sobeco!roe

dave@imax.uucp (Dave Martindale) (04/11/90)

I'll put in another vote of confidence for the Exabyte drives.

I have two of them.  I do use them for backups, but the main reason
I have them is to store images.  At 36 Mb per frame, and 24 frames per
second, that's a lot of bytes per second of animation.  Even the 240 Kb/sec
that the Exabyte provides is really rather slow - that's 2.5 minutes to
read a single image.  The 180 Kb/s rate of a DAT drive would be almost
intolerable.  And storing only 1/2 as much per tape would require changing
tapes twice as often when processing.  Even if I had a jukebox, I'd
have to reload it twice as often.

They've been quite reliable; either one will read tapes written on the
other, plus tapes from two other off-site Exabytes that I've used.  I have
had problems exactly once: "write data" errors while writing a new tape,
and cleaning the drive seemed to fix that.

I hear Exabyte has announced a new drive that doubles the capacity of
a tape, doubling the transfer rate at the same time (probably by just
doubling the linear recording density).  I eagerly look forward to it.