IBM Mainframe Forum

Hello All,

Please let me know your suggestions/advise for the following scenario I have.

I have a huge VSAM file say about atleast 500 million records. Most of these records are used only in certain conditions.
So iam planning to archieve these records prior to a particular date range onto a tape dataset ( Data prior to last one year will be archieved which contributes to atleast 70 % of the volume). But i might require these records under certain conditions in my processing. I can establish the requirement of these records in the process. But now I have problems on how efficiently I can bring back these from tape and use it in my process. ( I can use these back in a sequential or VSAM file). I have a major time constraint here on how quick I bring these records back from the tape. So my requirements are to bring these records quickly and efficiently and use it in my process with in a given time span. Im not aware if any special processing can be made on the tape files or any utilities available that can accomplish these.

Please let me know if any ideas or suggestions. Thank you in advance.

Perhaps you should also mention the kind of data set -- ESDS, RRDS, KSDS -- as this will of course make a major difference in recommendation.

Hi,

its KSDS !!!
Thanks.

Hello,

When these offloaded records are needed, can the process be run from a sequential file or do they have to be reloaded kinto VSAM?

Hi,
These records can be either into a sequential or VSAM file - my only requirement being efficient and quick retrieval from tape.

Thanks,

Hello,

Suggest you speak with your storage management people about migrating these records directly to offline storage (they would need to be sequential records rather than a "dumped" copy of the VSAM file - i'd use your sort product as it should be quite a bit faster than IDCAMS REPRO.

Just naming the migrated file in a batch job will recall it to dasd - which should be rather quick. . .

If you use sort to create your output try and sort so that the records most likely to be used are near the beginning of the file.

Why don't you split the file into several files, for example one file for each year? Then if no data is being accessed for a specific year the dataset will get migrated.