Joined: 23 Nov 2006 Posts: 19244 Location: Inside the Matrix
Some of the folks i work with have been given a new "opportunity" and have asked me for suggestions.
Currently, there is data (over 10,000 CYLs) they receive each month and for many years this has been copied to 3490 carts for long-term use (these are transaction history data and are periodically used to show all of the activity for a given account). Now, their boss has dictated that they must discontinue using 3490s and use "FAT" tapes (i don't know which model, but i don't believe this matters). They have also been directed to "stack" a year's worth of data on a single volume - this to save the cost of additional FAT volumes.
They asked about letting the disk files simply migrate and be recalled when needed and were told they cannot do this. Due to the volume of the data, they cannot use "virtual tape" either.
What would you suggest be done to "stack" a new month's files onto the same (or another) volume to join some number of previous month's data? Each January a new "stack" is going to be started.
The more i've thought about this, i don't recall there being a requirement to do exactly this anywhere i've been before.
One option i've thought about is to "toggle" between 2 volumes until year-end. Another consideration is that the dataset names have the year and month embedded in them.
My first recommendation was to simply use 1 FAT tape per month, but they said that was the first thing they suggested too and were told "no".
If this might be better served in some other part of the forum, it can surely be moved. . .
Joined: 28 Jul 2006 Posts: 1702 Location: Australia
Hi Dick,
the way I would do it is when it's Jan, create a new FAT tape, no need to toggle between 2 volumes and depending on the month, you run the relevant step
If you have a scheduling tool, it should not be too difficult to run the correct step based on which month of the year the job is running and build the correct date.
Joined: 14 Mar 2007 Posts: 8797 Location: Welsh Wales
I still prefer the HSM method. This is the scenario I employed for a similar situation where 9,000 cyls per month were created, and we had a legal requirement for 8 years retention. It depends on the requirements of data availability, but with the method below, it was belt, braces and thumbs in the belt loops.
1) A backup of each months data could be produced, according to MGMTCLAS, which offers additional security in case of media failure on the ML2 volume.
2) If there is media failure on the backup - then the backup can easily be recreated from ML2.
3) Even more secure is if the ML2 volumes are duplexed.
4) Using the MGMTCLAS for expiry, the tape usage released by dataset expiry can be reclaimed via RECYCLE.
Joined: 23 Nov 2006 Posts: 19244 Location: Inside the Matrix
Thanks for the replies,
Quote:
no need to toggle between 2 volumes
Maybe i am overly cautious - if the single volume is somehow damaged in August, the first 7 months data would be lost. . . Due to the size, the transmitted input data is deleted when copied. . . The "sender" is not required to keep a copy once the incoming data has been validated/verified.
Quote:
I still prefer the HSM method
Me too Unfortunately, they have been directed to "roll their own", so i've been trying to think of something usable, but painless