Re: CICS suffering from Page-In delay: Page Stealing versus Paging Out
Check SMF99 subtype 5 for storage target. SRM manages to a storage target which is set in a complex way if you are not using storage critical, but aims to keep interactive work responsive. Your CICS regions may be mis-categorized for this purpose. This storage target is enforced pre-emptively (even when there isn't a system-wide storage shortage yet), which is of course why this way of managing CICS regions has been replaced by storage critical. On Fri, Oct 9, 2015 at 10:34 PM, Peter Hunkeler wrote: > A word in advance: I have subscribed to the MXG-L list but have not yet > granted access, so I'm asking here first. > > > > I'm busy working on a problem some of CICS region are suffering from > periodically. There are times when some CICS regions are paging-in some MB > of storage and are thus suffering from the paging delay. > > > Since this is about page-in, the storage that is paged-out has been in use > before. It is not too clear yet what this storage is and what is causing it > to be referenced again in that massive way. That's being investigated and > is not directly my question here, although any hint is welcome as well. > > > I'm trying to understand why that amount of storage has been paged-out > from those CICS regions, and when this has happened. looking at the SMF30 > subtype 2 for the CICS regions a indeed can see intervals with large number > of pages being paged-out (SMF30PGO). I do not see any non-zero values > indicating pages had been stolen (SMF30PST). > > > > It has been my understanding (which may be wrong, of course) that page > steaqling happens when the Available Frame Queue goes below the min > threshold. SRM/RSM will them steal pages and page them out. I would expect > those numbers to appear in SMF30PST. And if this number is 0 for an address > space, this one was not a donor in a low AFQ situation, right? > > > > > We do see the AFQ going low sometimes, and it was my assumption that those > CICSs were victims by having pages stolen. This does not seem to be the > case. What then causes the pages to be paged-out (SMF30PGO)? Is CICS doing > PGOUT by itself? > > > > > > > -- > Peter Hunkeler > > > > > > > > > > -- > Peter Hunkeler > > -- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN > -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
AW: Re: CICS suffering from Page-In delay: Page Stealing versus Paging Out
> There is an option in the WLM to make a service class memory critical. Yes, I know and I have verified that we do *not* used it. Why? Good question that I will have to raise. I'm very astonished the CICSs are not set storage critical. But I'm new to this installation and don't know historical information available, yet. Apart from this, I have to live with the current setup and need to find answers to my previous questions. -- Peter Hunkeler -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: CICS suffering from Page-In delay: Page Stealing versus Paging Out
Check the storage critical value in WLM. This is the equivalent of the old "PWSS=" parameter in SRM. HTH, Al Staller | Z Systems Programmer | KBM Group | (Tel) 972 664-3565 | allan.stal...@kbmg.com -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Peter Hunkeler Sent: Friday, October 09, 2015 6:35 AM To: IBM-MAIN@LISTSERV.UA.EDU Subject: CICS suffering from Page-In delay: Page Stealing versus Paging Out A word in advance: I have subscribed to the MXG-L list but have not yet granted access, so I'm asking here first. I'm busy working on a problem some of CICS region are suffering from periodically. There are times when some CICS regions are paging-in some MB of storage and are thus suffering from the paging delay. Since this is about page-in, the storage that is paged-out has been in use before. It is not too clear yet what this storage is and what is causing it to be referenced again in that massive way. That's being investigated and is not directly my question here, although any hint is welcome as well. I'm trying to understand why that amount of storage has been paged-out from those CICS regions, and when this has happened. looking at the SMF30 subtype 2 for the CICS regions a indeed can see intervals with large number of pages being paged-out (SMF30PGO). I do not see any non-zero values indicating pages had been stolen (SMF30PST). It has been my understanding (which may be wrong, of course) that page steaqling happens when the Available Frame Queue goes below the min threshold. SRM/RSM will them steal pages and page them out. I would expect those numbers to appear in SMF30PST. And if this number is 0 for an address space, this one was not a donor in a low AFQ situation, right? We do see the AFQ going low sometimes, and it was my assumption that those CICSs were victims by having pages stolen. This does not seem to be the case. What then causes the pages to be paged-out (SMF30PGO)? Is CICS doing PGOUT by itself? -- Peter Hunkeler -- Peter Hunkeler -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: CICS suffering from Page-In delay: Page Stealing versus Paging Out
There is an option in the WLM to make a service class memory critical. It was designed especially for this issue with CICS. If CICS is not requesting the page out, this will protect all the regions in the service class. - -teD - Original Message From: Peter Hunkeler Sent: Friday, October 9, 2015 07:35 To: IBM-MAIN@LISTSERV.UA.EDU Reply To: IBM Mainframe Discussion List Subject: CICS suffering from Page-In delay: Page Stealing versus Paging Out A word in advance: I have subscribed to the MXG-L list but have not yet granted access, so I'm asking here first. I'm busy working on a problem some of CICS region are suffering from periodically. There are times when some CICS regions are paging-in some MB of storage and are thus suffering from the paging delay. Since this is about page-in, the storage that is paged-out has been in use before. It is not too clear yet what this storage is and what is causing it to be referenced again in that massive way. That's being investigated and is not directly my question here, although any hint is welcome as well. I'm trying to understand why that amount of storage has been paged-out from those CICS regions, and when this has happened. looking at the SMF30 subtype 2 for the CICS regions a indeed can see intervals with large number of pages being paged-out (SMF30PGO). I do not see any non-zero values indicating pages had been stolen (SMF30PST). It has been my understanding (which may be wrong, of course) that page steaqling happens when the Available Frame Queue goes below the min threshold. SRM/RSM will them steal pages and page them out. I would expect those numbers to appear in SMF30PST. And if this number is 0 for an address space, this one was not a donor in a low AFQ situation, right? We do see the AFQ going low sometimes, and it was my assumption that those CICSs were victims by having pages stolen. This does not seem to be the case. What then causes the pages to be paged-out (SMF30PGO)? Is CICS doing PGOUT by itself? -- Peter Hunkeler -- Peter Hunkeler -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
CICS suffering from Page-In delay: Page Stealing versus Paging Out
A word in advance: I have subscribed to the MXG-L list but have not yet granted access, so I'm asking here first. I'm busy working on a problem some of CICS region are suffering from periodically. There are times when some CICS regions are paging-in some MB of storage and are thus suffering from the paging delay. Since this is about page-in, the storage that is paged-out has been in use before. It is not too clear yet what this storage is and what is causing it to be referenced again in that massive way. That's being investigated and is not directly my question here, although any hint is welcome as well. I'm trying to understand why that amount of storage has been paged-out from those CICS regions, and when this has happened. looking at the SMF30 subtype 2 for the CICS regions a indeed can see intervals with large number of pages being paged-out (SMF30PGO). I do not see any non-zero values indicating pages had been stolen (SMF30PST). It has been my understanding (which may be wrong, of course) that page steaqling happens when the Available Frame Queue goes below the min threshold. SRM/RSM will them steal pages and page them out. I would expect those numbers to appear in SMF30PST. And if this number is 0 for an address space, this one was not a donor in a low AFQ situation, right? We do see the AFQ going low sometimes, and it was my assumption that those CICSs were victims by having pages stolen. This does not seem to be the case. What then causes the pages to be paged-out (SMF30PGO)? Is CICS doing PGOUT by itself? -- Peter Hunkeler -- Peter Hunkeler -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN