McKown, John wrote:
OK, I'm gonna ask this here. We have a 3590J tape subsystem. We have the
large, 300 GiB native, carts. Due to the time that it would take to fill
one up and restore the contents at D.R., we only really use about 1/4
the capacity. Does anybody use one of these to full capacity? If so, for
what? The thought of how long a restore would take of the entire
contents just staggers me. Oh, we also use them for our VTS backstore.
So I guess that it actually uses most of the capacity. Is anybody else
paranoid about what happens if one of these monsters snaps or is
otherwise rendered unreadable? We currently duplex all of our off-site
backups due to this and the possibility of damage during transport.
ITYM 3592J tapes (not 3590).
We use it, at full capacity.
Time to restore? As usually, IT DEPENDS.
If you do full volume dumps, or other "bulk" data transfer, the capacity
is not your enemy. You have to restore bunch of volumes or all the
datasets, so the data is probably read sequentially. Sequential reading
from Jaguar is *fast*. Faster than your DASD. The parameters we could
worry are mount time and positioning time, oftenly composed in "time to
first byte" value.
ML2 purposes, frequent, random reads/writes - that's bad scenario for
such tapes. In such case I would suggest STK/SUN 9840C - much faster
when considering time to first byte. Limited capacity of cart, worse
speed measuerd in MB/s. Or VTS/VSM solution - even worse speed, but
statistically 0 s time to first byte.
So, for DR purposes I would be happy to have 3592 or T10000. For ML2
and/or batch directly using tape - I'd woory about huge tapes.
Last but not least: Lost tapes. IMHO the pain is irrelevant to cartridge
capacity. What would you do if you lose only 1GB of your database (but
let Mr. Murphy choose WHICH ONE !) ??? I know what I would do: use
another copy. I duplex *all production data* on tapes.
I can't remeber when I sufficed media error last time. However I was
told it still use to happen <g>
I prefer to hear horror stories, not to experience them.
Regards
--
Radoslaw Skorupka
Lodz, Poland
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
