Ron and Jenny Hawkins wrote:
John,
You can take a single table and make it multiple physical files
(partitions) and backup those files at the same time. Say you have one
table that is 40 GB, but you spread it across 10 files of 4GB each,
spread across 10 different DASD volumes (disk drives). Instead of
backing up one 40 GB file, you can now backup ten 4GB files concurrently
(assuming you have 10 tape drives). No your backups run in 1/10 of the
time. This is where the tape system comes into play, instead of a human
taking 5 minutes to mount each tape (50 minutes total) a tape system
could have all 10 mounted in 5-10 minutes, a 40 minute savings in time
to get your backups done.
40GB! Like who cares about 40GB. That's a typically small LUN (volume) in
Unix and Wintel (my PC hard Drive is bigger than that). Using Logical Volume
Manager RAID-0 a couple of TB of Oracle Database tables are striped across
many LUNs, and these are backed up to tape with a degree of parallelism that
would make most mainframe shops green with envy. We have Unix shops with
Disk to Tape backup throughput requirements of 1-2GB/sec. Compare that to a
maxed out VTS that struggles to get up to 400MB/sec.
Although this was just an example to show you can have multiple physical
files as one table space and backup each partition concurrently, it is
not that far off from the truth, at least under z/OS. Largest DASD
volume that z/OS supports is about 25 GB and IIRC DB2 supports physical
files up to 16 GB, but you can have 4096 of them in a single partitioned
table space.
Umm, 1-2GB/sec of DASD to Tape, that fast. Can you point me to a
reference for this? I had been under the impression that most tape
units can only do 30-60MB ps. IBM and Sun do have tape units that can
do about 100MBps.
All robotic tape system of any size will cost millions of dollars, not
matter whose it is. However, they save 10's of millions of dollars in
time and staffing, at least they are supposed to.
Interesting, millions of dollars for a 10 drive library. We'll have to tell
ADIC, Quantum, IBM and Stortek to increase their list prices. Surprising as
it may seem, those 100s of servers can get by without a Powderhorn class
library. And as for 10's of millions of dollars of time and staffing, well
that's an interesting argument in Asia where you can hire an educated and
intelligent carbon based robots for around US$2K a month.
I guess I was not clear. The 10 drive library was part of my example
about backup up a single table space that is partitioned into 10
partitions can be backup concurrently. It was "as400" that had the
multi-million dollar VTS and I have no idea how many tapes drives where
in that unit.
I would assume that if it cost a million dollars from IBM that the same
type of VTS would cost about million dollars from somebody else.
If a VTS is not supposed to save money due to staffing and time, then
why get one?
Ron
----------------------------------------------------------------------
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
