ADSM-L

Re: Tape Reliability Recommendations

2003-02-19 14:03:12
Subject: Re: Tape Reliability Recommendations
From: "Prather, Wanda" <Wanda.Prather AT JHUAPL DOT EDU>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Wed, 19 Feb 2003 14:02:52 -0500
Adding a bit of my own experience to Kelley's:

Even though we are all using TSM, we use the hardware differently.

At one time we had two DLT libraries, with libraries and drives provided by
the same vendor.
Identical hardware, manufacturer, media, and microcode, same level of TSM
running each.

One we got perfect reliability, the other was a nightmare - constant read
and write errors, and many drive failures.  We couldn't stop the problems,
even after replacing all the drives, more than once.  After a LOT of time
spent with the vendor and hardware gurus, we finally learned:

Part of it is just the total load on the hardware, and the other part is the
TYPE of backups you run.

- If you are doing a LOT OF TINY files - for example, workstation/desktop
backups - you will get a tremendous amount of start/stop activity (or call
it backhitch, or repositioning, whatever) during migration to the tape, and
during reclaims.  TSM uses tape almost like a direct access device, and this
pushes the media and the drive to their max capability.  You need the best
drive mechanics and the best media you can buy.  And you will need to be on
GOOD TERMS with your vendor and stay on top of those microcode levels.

-If you are dumping a FEW BIG FILES daily - for example, huge databases -
you tend to write one or two big files on the tape and you're done.  The
tape sits in your vault for a while, then you do the same thing to the tape
again.  Even though you may be writing MORE GB PER DAY than with the
workstation model above, it's far less stress on the drives and the media.
You will probably get better reliability on your drives/media than someone
doing workstation backups with the same hardware.

My opinion and nobody else's,
Wanda Prather





-----Original Message-----
From: Kelly J. Lipp [mailto:lipp AT STORSOL DOT COM <mailto:lipp AT STORSOL DOT 
COM> ]
Sent: Tuesday, February 18, 2003 1:32 PM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Re: Tape Reliability Recommendations


I have done a significant amount of testing and have quite a lot of
practical experience with what I will refer to as the Big Three tape
technologies:

AIT3
LTO1
SDLT320

Of the three, I know AIT the best.  It's always good to know where someone
sits before they tell you where they stand.

In a TSM environment, all three of these technologies perform very
similarly: within 10-15% of each other.  Don't let the manufacturer's
performance claims sway your decision.  Backup is generally not about the
hardware but more about the software.  TSM is quite powerful, but often
trades power for speed.  Remember, we have a sophisticated database running
here to track what's going on.

For all three drives, we are able to sustain between 35 and 45 GBytes per
hour during storage pool to storage pool operations.  For instance,
migration from Disk to Tape or Backup stg tape to tape.  In addition, you
can expect to see about the same performance when clients are writing data
directly to tape (or even multiple tapes simultaneously while using the stg
pool parameter copystgpool).  When sizing an environment, use the 35 GB/Hour
number and you won't be unhappy.

As for reliability.  That turns out to be a very mixed bag.  I have seen
sites with high volumes of data and no errors or problems with all three and
I have seen sites with numerous problems.  The problems seem to be mostly
related to drive firmware levels and tape batches.  Once the drive firmware
is correct and bad tapes are eliminated, most sites settle down nicely.  The
more complex the environment, the more likely the problems.  For instance,
we have a site with a large fiber channel and LTO configuration.  No end to
the problems so far and they are very serious problems.  Is this a result of
the tape technology?  I doubt it, but one never knows, do one?\

Due to the nature of AIT3, I would suspect that overall reliability numbers
will be lower than for LTO and SDLT, but my hands-on experience doesn't show
that.

As for Automation.  There are gazillions of libraries for each technology.
Clear winners in my opinion are Qualstar and perhaps IBM.  I give the IBM
libraries a "perhaps" as we have had very good experience with the 349x
libraries and only limited experience with 3584.  These seem OK, but not
much experience.  The lower end IBM libraries are based on someone else's
technology so I would think one might get a better deal buying direct from
that manufacturer.

Compatibility with previous technology.  Some DLT bigots are SDLT bigots
because they believe in investment protection.  I think that's balderdash as
very few people would ever try to read a DLT tape with an SDLT drive anyway
so what difference does it make?  All three of these are relatively new
technologies and you are going to switch to one anyway, so investigate all
three.

The all important Kelly recommendation:

For value, AIT3 is unsurpassed: very good performance, relatively
inexpensive, great automation, manufactured by one company so technology is
first rate.

For openness (or perceived openness) LTO: excellent performance, reasonably
priced, so-so automation, standards based and built by more than one
manufacturer (but how many of us are going to buy from more than one anyway
and if you attend presentations by each one about their LTO product you come
away from each one in succession thinking you have found the best, i.e.,
they all lie equally convincingly (probably shouldn't have two "ly" words in
the same sentence)).

For perceived technical excellence, SDLT: Quantum has very neat technology
in their drives.  Does it matter much?  Probably not, but cool anyway.

So:

For the price conscious: AIT3 going to AIT4 when available.
If you're an Open kind of dude: LTO
If you believe in Quantum: SDLT.  They offer a very good product IMHO.

LTO and SDLT will be very close in price so go with your gut.

As always, study, study, study.  Get input from those you respect.  Choose
wisely and then get and stay behind your choice.  STORServer supports all
three technologies equally.

Views expressed here are my own.

Kelly J. Lipp
STORServer, Inc.
485-B Elkton Drive
Colorado Springs, CO 80907
lipp AT storsol DOT com or kelly.lipp AT storserver DOT com
www.storsol.com <www.storsol.com>  or www.storserver.com
<www.storserver.com>
(719)531-5926
Fax: (240)539-7175


-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU]On
<mailto:ADSM-L AT VM.MARIST DOT EDU]On>  Behalf Of
Colby Morgan
Sent: Thursday, February 13, 2003 3:24 PM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Tape Reliability Recommendations


We are currently running TSM 5.1.5 on Win2k with an IBM Mammoth-2 drive for
offsite copypools.  We have had problems with both our onsite M2 and offsite
M2 drives at our disaster recovery center.  IBM has replaced the drive more
than a dozen times in the last two years and Exabyte has replaced countless
tapes.  Most recently we are experiencing a high rate of media write
failures on a newly replaced drive as well as media read failures in DR
testing, both using brand new 225m AME media.

Is anybody else out their running an IBM/Exabyte Mammoth-2 drive and if so
what kind of results do you see?

My real question is what is the most common/reliable removable tape
technologies for the Intel TSM environment?  We are considering switching
technologies and I wanted to solicit testimonies on other technologies (DLT,
LTO, SDLT,etc...).

We currently copy around 135GB to 300GB offsite daily.


Thanks,

Colby

<Prev in Thread] Current Thread [Next in Thread>