I don't expect below-the-line CSA to be much affected by a hardware change (not in 2018 for sure) but in the case of a true push/pull swap out, you are likely using a whole new IODF--which is software. That was certainly true for us. Putting aside the question of whether this explains the current problem, let me share the advice I always give my colleagues.
As Peter notes, the dividing line between common and private is determined by the system at IPL time. I don't know all the details, but the process is something like this. 1. Calculate the 24 bit PLPA total. 2. Add in the amount of 24 bit common requested in PARMLIB member CSA=. 3. Round up to the next 1M boundary. Everything below that boundary is 24 bit COMMON. Everything above it to the 16M line is 24 bit PRIVATE. You can tweak the contents of 24 bit PLPA to some extent, but the main control is CSA=. Increasing that value raises PRIVATE at the expense of COMMON, and vice versa. Changing the ratio one way or the other can be problematic, especially if it's unexpected. For example, CICS tends to want lots of PRIVATE, while DB2 tends to want LOTS of COMMON. (At one time anyway.) So you want to keep a ratio that is known to work. My advice is to look at your happily running system. Calculate a value in the middle of the 24 bit COMMON area; basically 500K below the boundary. That is the sweet value for CSA=. It allows for maximum fluctuation up or down in PLPA before the boundary changes unexpectedly. You need to review this calculation periodically, but don't obsess over it. . . J.O.Skip Robinson Southern California Edison Company Electric Dragon Team Paddler SHARE MVS Program Co-Manager 323-715-0595 Mobile 626-543-6132 Office ⇐=== NEW [email protected] -----Original Message----- From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf Of Peter Hunkeler Sent: Tuesday, January 08, 2019 11:03 PM To: [email protected] Subject: (External):AW: Re: Generic query on Region allocation failure >So when we move to a higher version of hardware the storage area below >the line shrinks ? Not necessarily. It may or may not, and the cautious system programmer may or may not be able to avoid it. There are certain boundaries in the address space map that must lie on a megabyte boundary. The boundary between the common and the private areas is one. A single byte more used in the common area might move the common-private boundary down by one megabyte. One is often able to reduced the size of some common areas (CSA/SQA) by a small about and thus avoid the common-private boundary shift. HTH Peter Hunkeler ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
