[rescue] Re: Quick EXB-210 questions

[Greg A. Woods]

[ On Monday, August 27, 2001 at 14:27:23 (-0700), Robert Novak wrote: ]
> Subject: [rescue] Re: Quick EXB-210 questions
>
> MP, AME, and AIT physical media are different.

I wasn't talking about the media as a package -- I was talking about the
tape and emulsion itself.

Yes, MP is different from AME or AMP.

>From what I know there doesn't seem to be any huge difference between
AME and AMP

AIT cartridges contain, if I'm not mistaken, AME tape.

Super-DLT has AMP, but with optical servo info on the back of the tape.

> According to several
> sources the tape life (years and passes) are about the same.

Yes, that's probably true, but you're ignoring the fact that any
generation of DLT tape life exceeds anything used by any generation of
helical scan technology by well over an order of magnitude in terms of
passes; and probably significantly more real-world shelf life in terms
of longevity too!

Just look at the tape path and the number of moving elements it must
pass over in any helical scan device vs. DLT and you can't help but
notice the reasons why DLT tape life is longer in all ways.

> In real life
> settings, I got about 4MB/sec/drive on Mammoth and 10MB/sec/drive on
> AIT-2, each with two drives on the same bus, and with a 50/50 mix of local
> and network storage going to the drive. This was based on 2 years with the
> Mammoth and 1 year with the AIT-2 changers/drives. 

I think you're comparing different generations of Mammoth and AIT.

Mammoth == AIT-1

Mammoth-2 == AIT-2

In comparison with any helical-scan technology it seems DLT has always
been a generation ahead in terms of tape write speed though....

> This is not accurate from any real world setting I'm aware of, unless
> there is some element of "full advantage" which specifically specifies
> "works with such-and-such software." I used Legato Networker and Solaris
> ufsdump and they both made quite substantial use of the AIT2 drive and
> media. Same with Mammoth 1. 

What's the "AI" in "AIT" stand for again then?  Maybe Legato supports
the MIC, and mabye Solaris does too, but I'm not so sure.

> I am not aware of any magic in the AIT world--it just works. I got a

You're using "AIT" as if it's just "T" then....  Why pay all the extra
money, and "suffer" the lower duty cycles, then?

> consistent 125-130GB/tape (50GB native) using the hardware compression as
> /dev/rmt/#hbn with no driver hacks or special software. This was only
> observed over the course of 11 months, with about 300GB of data 50/50
> local and network, so it may have been an anomaly. Your mileage may vary.

As for data compression in the drive, well I'd never use it or recommend
it, though for some folks with some types of data there can be
advantages....  Anyone comparing throughput with compression on is
proably slanting their data in ways more favourable to their algorithms.

(Sony licensed IBM's addaptive lossless compression algorithms, which
gives them favourable compression ratios (up to 2.6-to-1) and speeds in
comparison to DLT* and Mammoth* (only up to 2-to-1), but that point is
moot to me.)

The only meaningful numbers are the raw read&write rates.  The only
thing to worry about, if you need drive compression, is to make sure the
drive never writes slower when compressing than it can without
compression (as almost always they'll go faster -- treat the "faster" as
gravy and always plan on raw uncompressed rates as your baseline.  Never
expect any compression or any increase in speed because of compression.

> What I really wonder though, is if you're saying AIT costs more by volume
> (i.e. $/GB) than Mammoth or DLT? If so, back it up with some real prices
> and real world compression rates, not list price and published rates.

I think the documents I posted links to can do that just fine....

Mammoth-1 and DLT4000 compare pretty equally on TCO.  I've not done the
research and math, but I'll bet Mammoth-2 and DLT8000 compare similarly
on TCO.  I can't do a current AIT real-world TCO comparison because none
of the dealers I buy magnetic media from even stock Sony's AIT or AIT-2
(though one of them will sell me a drive! ;-)

As for raw write speed, these are the number's I've collected
(translated to a common KB/s):

4mm DDS 	 183 KB/s
4mm DDS-2	 510 KB/s
4mm DDS-3	1000 KB/s

8mm 8205XL	 270 KB/s
8mm 8505XL	 500 KB/s
8mm Mammoth	3000 KB/s

8mm Sony AIT	3000 KB/s

Tandberg MLR1	1500 KB/s

DLT2000XT	1250 KB/s
DLT4000  	1500 KB/s
DLT7000  	5000 KB/s

Sorry, I don't have real-world numbers for Mammoth-2, DLT8000, or AIT-2,
but:

AIT-2 is supposedly 6000 KB/s

SuperDLT is supposedly 11000 KB/s

Rumour has it that AIT-3 will catch up again with 12000 KB/s.

Mammoth-3 is on the way too.....

> > Exabyte's SCSI mode pages aren't exactly 100% standard either, but at
> > least they're fully and publicly documented and can be used without
> > major driver hacks.  If you want a machine readable label for your
> > robot Exabyte supports a standard SCSI bar-code reader in their
> > changers -- heck even Amanda supports that!
> 
> Well, this one doesn't make a lot of sense. I've used Qualstar and Exabyte
> Mammoth libraries, and Overland Data and (soon) ADIC AIT2 libraries, and
> they ALL offer barcode readers which are supported under conventional
> software packages.

Read what I said again, but this time keeping in mind what the "AI" in
"AIT" stands for, and what it means from the engineering side.

Not only that but an externally readable bar code has many advantages
over the proprietary machine-only readable MIC in AIT for storage of any
"user-definable" items!

Everything I've read by third-party reviewers consistently says that
Sony AIT's only advantage is the MIC, but from what I can see that's a
very slim advantage and a very quick way to get locked into something
far more proprietary and expensive than the rest of the bunch.

Many of the reviewers seem to ignore the fact that modern DLT drives
also store similar internal-only indexes and partition tables at the
beginning of the tape too (as do Exabyte's too IIRC).  The load time of
this information in comparison to the total tape operation is nearly
invisible unless you're trying to find one file among thousands of tapes
(in which case the MIC could make the difference in not needing any
physical tape movement).

Why buy a 64KB flash-ram chip in every cartridge (plus the electronics
to read and write to it) when you've got oodles of equally good magnetic
media sitting right there too?!?!?!?!?

> Exabyte even has AIT these days, by the way, alongside 
> DLT and M2.

Some people will do anything to gain market share.  Exabyte sells all
kinds of personal/PC backup stuff too.  Their DLT and AIT offerings are
solely for pushing their libraries and autoloaders out into a wider
market though, and all they do is buy the other guy's tape drives and
stick them in their libraries.  They don't license or manufacture either
of those technologies.

>From a tape drive perspective they're now again almost entirely focused,
on their own line of Mammoth (and Eliant last of the old 8mm) devices.

I'm not a big fan of Exabyte -- I never was (I got a bad start with them
a LONG time ago) -- though I do admit they make great tape library
robots!  (The very best mid-sized one for DLT, IMNSHO, and no doubt for
AIT too.)

I'm just far less of a fan of Sony.  They are supreme experts that the
"proprietary lock-in" game (while at the same time marketing media for
everyone else's drives!  ;-).

After all, why is it that all the vendors I buy media from offer all
forms of Sony 4mm, 8mm, and DLT (not to mention diskettes and CD-R and
CD-RW) media (as well as Fuji Maxell, Exabyte, IBM, etc.) while not
carrying Sony AIT media?

> I don't see the existence of a standard barcode reader as a good reason to
> choose one medium over another. Kinda like choosing one car over another
> because you can get fuzzy dice for one of them. 

The point is/was that if you want more portable support for machine
readable tape labels and indexes then get a standard bar code reader!
Don't put a stupid proprietary flash-ram chip in every friggin' tape
cartridge!

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods at acm.org>     <woods at robohack.ca>
Planix, Inc. <woods at planix.com>;   Secrets of the Weird <woods at weird.com>