On Tue, 8 Dec 2009 08:29:36 -0600, Arthur Gutowski <[email protected]> wrote:
>On Mon, 7 Dec 2009 11:43:12 -0600, Mark Zelden ><[email protected]> wrote: > >>The LDA contains what the sizes are, but I've never updated those >>control blocks directly via IEFUSI. >> >>FWIW, I look at the GDA and for below 16M set region limit to "ALL-512K" >and >>region size to "ALL-64K". There is no real reason to reserve LSQA above. >>Whatever isn't in use for extended private can be used for LSQA. When >>those 2 boundaries hit each other is when your address space will die. > >Mark, thanks for chiming in - I'm glad I held off. I should probably not have to >ask this, but programming IEFUSI to reserve ELSQA was something ingrained >long ago, and I just took it for granted. > >So, it's not what the REGION is, it's what is GETMAIN'd in total during >execution? I ask because we recently decided to bump the ELSQA reserve for >our CICS regions, but in light of this, I don't see any benefit. I know CICS >manages its own memory (DSA/EDSA), though I'm not as familiar with *how*. > >Interesting how different this is from below-the-line... > >By the way, I had an ETR discussion quite some time ago with IBM (I may be >able to find it in my archives), in which I recall we're not supposed to alter >LDA. Like I said, I've never done it that way and didn't know anyone else did. That is not the way the exit interface was intended as far as I can tell. The way I look at it, LDA control blocks are the "result" of the processing that takes place after IEALIMIT / IEFUSI. I wouldn't have even known if changing the LDA directly would have the desired affect without testing (I assume it does though). > In fact, that's what I use to obtain available REGION. I thought about >using GDA, but do not recall why the rep steered me away from it... > >Though LDAESIZ will not completely protect against EPRIVAT and ELSQA >running into each other, we do use it to try to buffer against the large VL >GETMAIN. Do you do this, or is it "let the user beware"? Just curious here... > I still prevent REGION=0K/0M from all but STCs. This really is a moot point on all but a couple of very small monoplex LPARs. With the next smallest LPARs having about 9G of real storage and very robust paging subsystems, I don't think I have to worry about it. :-) If you recall some past posts of mine, even though I still prevent REGION=0K from batch jobs, I allow REGION= whatever you code - and if you code anything over 1024M, I give you everything above 16M, so you can get the equivalent of REGION=0M above the line anyway (I did that in 2004). Just not by coding REGION=0M. I still left in the STC check for REGION=0K/0M because that also allows for MEMLIMIT=NOLIMIT (again, I only allow that for STCs). Mark -- Mark Zelden Sr. Software and Systems Architect - z/OS Team Lead Zurich North America / Farmers Insurance Group - ZFUS G-ITO mailto:[email protected] z/OS Systems Programming expert at http://expertanswercenter.techtarget.com/ Mark's MVS Utilities: http://home.flash.net/~mzelden/mvsutil.html ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
