04-14-2016 03:08 AM
When we did this kind of migration (z/OS to AIX), the biggest hurdle was the organisation of files/libraries.
On the mainframe, the historical way was to have one library per dataset, because of the way the datacenter people wanted to handle their DSNs in the JCLs.
When we migrated, we set up the first qualifier of the mainframe partitioned datasets as a library definition (and directory) in UNIX, and put multiple datasets into those libraries
XYXY.SAS.DATA1512 (z/OS partitioned dataset containing SAS dataset DATA1512)
/corporatedata/xyxy/data1512.sas7bdat, with library XYXY assigned globally (autoexec_usermods) to /corporatedata/xyxy
So the end users had to rewrite their dataset names in set statements, but could completely omit all the necessary libname definitions they had to do earlier.
Since the first qualifier of the z/OS dataset names was also used for access rights, this logical construct could then be applied on the UNIX directories (/corporatedata/xyxy and all datasets therein belong to group xyxy)
Codes could mostly be taken over without changes apart from those relating to dataset names.
Be aware that in EBCDIC lower case letters come before uppercase (ASCII is exactly the opposite) and that the sequence of characters is also different ('Z' and 'A' are 25 characters apart in ASCII, but 40 in EBCDIC). So everything that relies on the ordering of characters in the encoding needs to be revised.
Next question is how you import data from production data sources into SAS. If you used plain text, the file transfer from z/OS to UNIX can take care of that. If you have packed decimal data, you need to transfer in binary mode and use S370 and EBCDIC informats.
04-14-2016 06:01 AM
We have no documents any more. You see, this was done here sometime before 2000, with SAS 6.9 on the MF and 6.11 on Reliant UNIX, with a subsequent migration to 6.12 on AIX.
What I told you comes from the back of my head and some consulting of wikipedia (regarding peculiarities of the EBCDIC character set).