FWIW, adding an additional STORAGE frame to you 3494 is probably your
least-cost option in the short term. D-frames (drive) are expensive, but
S-frames (storage) are really cheap - you will spend more on media to fill
it than on the frame!
We are in somewhat the same position as you. We aren't concerned about our
media life, but we need additional capacity. As soon as the $ are available
we have decided to go ahead and start converting our TSM from 3590 to LTO.
Not that we are unhappy with the 3590, but for future growth, we can't
justify the cost of 3590-H or 3592 IN OUR ENVIRONMENT.
No one, including IBM, has EVER said that the LTO drives are equivalent to
3590 (or STK 9840's) in terms of reliability, speed, and duty cycle. I have
no hesitation at all in running my 3590's or 9840's doing reclaims 24 hours
a day, if necessary. In terms of reliability of drives and media, I think
they are the best thing you can buy (except maybe those things they put in
airline black boxes :>). We found we could NOT run our DLT drives nearly
that hard, we just shredded both drives & media. LTO - so far we have had
NO problems at all, but I don't think anybody has had LTO long enough to see
what type of drive reliability you get when the drives start to age and you
pound them really hard.
Our TSM enviornment is not particularly dynamic; we don't get a lot of
user-generated ad-hoc calls for quick restores on our server that uses
3590's. As such, LTO meets our TSM requirements. And for future growth,
there is just no way we can justify the cost of the 3590-H or 3592 when we
can get the capacity and satisfactory (if not optimal) speed with LTO so
cheaply. (However, if this were an HSM environment, that would change the
equation a lot.)
So in short, consider your environment. If your concern is big bulk, go
with LTO. If you have a need for speed, stick with 359x.
You are probably uniquely positioned to TEST it yourself, since you already
have both types of hardware.
What is your worst fear? Having to restore one huge server in the minimal
amount of time, or respond to a dozen requests for single files?
Set up a test machine and run the restore. Then mark all your ONSITE tapes
as UNAVAILABLE, and run the restore again - that will force the restore to
come from your LTO copy pool instead. See what you like and don't like
about the results. LTO may be sufficient, maybe not, for your needs.
"I/O, I/O, It's all about I/O" -(me)
From: James R Owen [mailto:Jim.Owen AT YALE DOT EDU]
Sent: Monday, April 05, 2004 4:39 PM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: ?continue using 3590's or convert to LTO tape?
To other TSM sites w/ substantial investment in 3590 technology:
We are again trying to figure out what to do for additional TSM tape
capacity. We currently have two IBM 3494 ATL's (each w/ six 3590E1A
SCSI-connected drives and ~2000 tapes in each) for primary STGpools.
We also have two IBM 3584 ATL's (one w/ four LTO1 and two LTO2 drives,
and another w/ three LTO1 and three LTO2 drives, all FC and all
using LTO1 media.) [The LTO's were our remote, online Copy STGpools.]
To continue using 3590 media for our primary tape STGpools, we either
need to buy an additional 3494 frame (used?) or convert our existing
drives from 3590E's to 3590H's (used?). The E->H conversion would
eventually give us 50% more capacity with our existing media via TSM
tape reclamation processing. That seems reasonable to me, but...
Others here speculate that perhaps we should consider converting from
3590 to LTO media because of increasing 3590 maintenance costs, etc.
Another concern is that in 3-5 years our 3590 media may be obsolete
[definitely beyond the 10 year warranty period], so why not get off now?
Are you [other sites substantially invested in 3590's] continuing to
invest in 3590's -or- are you converting to other (LTO?) tape technology?
Is anyone considering/already doing a substantial 3590->LTO conversion?
If so, we would like to talk with you.
Thanks for your help.
jim.owen AT yale DOT edu (203.432.6693)