IBM Mainframe Forum

Ho hum...
This is not a simple subject, and you are trying to make it such - then adding "so will this be good from performance point of view".... 2 different sides of the coin
Real life :
DFSHDC40 is an excellent general purpose randomiser [but everything depends on your keys]
You can amend your parms to accommodate 500000 records [assume you mean records and not segments ] on the basis of 1 rap per record and given N raps/block arrive at an RAA [and DBDS size]
You can then allocate or [over allocate] the DBDS
It will load [if the dataset can grow ...overflow]
It will return records [just like SQL that is syntactically correct]....
...will it perform ? .... It "depends" - but we have done that above

If you are really interested in this sort of thing talk to your DBA [or do some more reading]

/OUT

Thanks dfshdc40!!. Yes, by 50000 records, I meant 50000 root segments(hence 50000 records). I feel, if hdc40 couldn't generate unique raps for all these unique 50000 root segments, then synoym chains will get generated, meaning 1 rap points to more than 1 record. thus performance may go down!!!!!

so looking at your comment, is it safe to derive now that only if requester(programmer) gives me 50000 unique key values for 50000 unique root segments, the only there is probability that hdc40 achieves magic of creating 50000 raps??. I am sure there is a way where we can prompt hdc40 to generate unique raps...for ex...1)providing unique input keys...

pls note there is no dba in my shop...I am the one who is doing this work and am completely new to ims db..

Regards
Panda15

"pls note there is no dba in my shop...I am the one who is doing this work and am completely new to ims db.."

Sorry to hear that you are between a rock and a hard place ... this is a management issue, not a tech issue - get trained or it's time for some discovery learning