steve-myers
Active Member
Joined: 30 Nov 2013 Posts: 917 Location: The Universe
|
|
|
|
Before - and I know this goes back a long, long way for all of us grey beards - using JCL PASS for a cataloged data set before VSAM BCS catalogs should have resulted in a small run time improvement. In other words, JCL like this -
Code: |
//A EXEC PGm=...
//ADS DD DISP=(SHR,PASS),DSN=...
//B EXEC PGm=...
//ADS DD DISP=(SHR,PASS),DSN=...
//C EXEC PGm=...
//ADS DD DISP=(SHR,PASS),DSN=...
//D EXEC PGm=...
//ADS DD DISP=(SHR,PASS),DSN=... |
should have run faster than JCL like this -
Code: |
//A EXEC PGM=...
//ADS DD DISP=SHR,DSN=...
//B EXEC PGM=...
//ADS DD DISP=SHR,DSN=...
//C EXEC PGM=...
//ADS DD DISP=SHR,DSN=...
//D EXEC PGM=...
//ADS DD DISP=SHR,DSN=... |
because the catalog lookup time was probably much slower than retrieving data set information from the JCL "passed data set queue."
These days, of course, catalog lookup times are much faster, though, of course, it is still, I think, slower than using the passed data set queue.
I'm curious, does anyone have a contrary thought about this - preferably backed up by hard numbers - than my long held opinion? Does buffering in the catalog address space affect lookup times. In my example there is no known impact as step run times increase, though that might be an issue with long step run times. |
|