Re: Paging subsystems in the era of bigass memory
z/OS MVS Initialization and Tuning Guide: Because SCM does not support persistence of data across IPLs, VIO data can only be paged out to DASD. Therefore, even when SCM is installed you must still maintain a minimum amount of storage that supports paging for all of your VIO data, and a minimum amount of local paging data sets. All other data types can be paged out to SCM. Jim Mulder z/OS Diagnosis, Design, Development, Test IBM Corp. Poughkeepsie NY > Since we upgraded to the z13 we increased real drastically on all > lpars. Recently, we also started using SCM (flash express). The > only thing we were baffled by was the fact that we still needed > *one* local page data set to satisfy the system. Other than that, > we're very happy with SCM (where we already have it active). -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
d10j...@us.ibm.com (Jim Mulder) writes: > it might be faster to read it from the DB2 data set, because DB2 > (via Media Manager) uses zHPF, but z/OS has not been enhanced to > use zHPF for page data sets. in 1980, I got con'ed into doing channel-extender for STL that was moving 300 people from the IMS group to offsite bldg. with service back to STL datacenter. It had channel emulator box at the offsite bldg with local channel attached 3270 controllers at the offsite bldg. there was full-duplex streaming protocol with the offsite channel emulator with channel programs downloaded ... significantly mitigated the latency of the heavy duty half-duplex channel protocol chatter. http://www.garlic.com/~lynn/submisc.html#channel.extendeer The vendor tries to get IBM to release my support ... but there was group in POK that was playing with some serial stuff that blocked it because they were afraid that it would make it more difficult to get their stuff released. In 1988, I'm asked to help LLNL get some stuff they are playing with standardized ... which quickly becomes fibre channel standard ... it includes full-duplex streaming (initially 1gbit/sec in both directions) and downloading i/o programs (countermeasure to protocol chatter latency). In 1990, the POK serial stuff is finally released as ESCON, when it is already obsolete. Later some POK people become involved in fibre channel standard and define a heavy weight, high latency protocol chatter protocol that significantly cuts native throughput that is eventually released as FICON http://www.garlic.com/~lynn/submisc.html#ficon Most recent published "peak I/O" benchmark I've seen is for z196 that got 2M IOPS with 104 FICON (running over 104 fibre channel standard). About same time there was a fibre channel standard announced for e5-2600 blade that claims over million IOPS (for single FCS, two would have higher native throughput than 104 FICON). zHPF w/TCW describes something a little like what I originally did back in 1980 ... but says it only provides 30% improvmeent over standard FICON (possibly only 70 FICON to get 2M IOPS, compared to only two fiber channel getting over native 2M IOPS). This recent post about 1987 (IBM) disk "wide-head" proposal capable of something like 48-67mbytes/sec transfer ... by 1990, FCS would have over 100mbyte/sec (1gibt/sec) full-duplex. At the time standard IBM had 3mbyte/sec channel moving to 4.5mbyte/sec ... and ESCON in 1990 would only be be 17mbytes/sec ... so as an IBM product, "wide-head" wasn't feasible. http://www.garlic.com/~lynn/2017d.html#54 GREAT presentation on the history of the mainframe http://www.garlic.com/~lynn/2017d.html#60 Optimizing the Hard Disk Directly recent paging subsystem posts here http://www.garlic.com/~lynn/2017d.html#58 Paging subsystems in the era of bigass memory http://www.garlic.com/~lynn/2017d.html#61 Paging subsystems in the era of bigass memory http://www.garlic.com/~lynn/2017d.html#63 Paging subsystems in the era of bigass memory http://www.garlic.com/~lynn/2017d.html#65 Paging subsystems in the era of bigass memory http://www.garlic.com/~lynn/2017d.html#66 Paging subsystems in the era of bigass memory -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
And if the buffer page is paged out to SCM, it should be faster to read it from SCM than from the DB2 data set. But if it is paged out to a page data set, it might be faster to read it from the DB2 data set, because DB2 (via Media Manager) uses zHPF, but z/OS has not been enhanced to use zHPF for page data sets. Jim Mulder z/OS Diagnosis, Design, Development, Test IBM Corp. Poughkeepsie NY > 30 years ago this type of logic, to reread from DASD rather than demand > page-in a buffer page, *might* have made sense, when the I/O bandwidth > to the paging subsystem was limited based on the number of page datasets > and a demand page rate of 30-100 pages a second was common to see. But > the world has changed and it would not make sense to do it today. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On Tue, 18 Apr 2017 00:47:55 -0500, Barbara Nitz wrote: >>I don't want to turn this into a debugging session, but this is what I see in >>RMF3 on the non-DB2 utility system. >>All of the TPX 'regions' are below this list sorted by Aux Slots. No task >>shows any page-ins. > >Not sure that you haven't already done this: Use IPCS active on that system, >and use either ip ilrslotc or panel 2.6i. It'll show you exactly which address >space paged out anything to aux (to whatever medium). >I recall that in a prior life I could see exactly which asid had a lot of >frames on aux (MQ, I think). > I've been using this REXX for years instead of getting into a monitor or IPCS. Very quick to execute and see the results. It was posted to the IBM-MAIN newsgroup only, so it's not in the archives. To browse the results I execute it using one of my execs to capture TSO output instead of "TSO %execname". https://groups.google.com/d/msg/bit.listserv.ibm-main/bVaW_PGzbaI/Wut56jByETMJ Best regards / Mit freundlichen Grüßen, Mark -- Mark Zelden - Zelden Consulting Services - z/OS, OS/390 and MVS ITIL v3 Foundation Certified mailto:m...@mzelden.com Mark's MVS Utilities: http://www.mzelden.com/mvsutil.html Systems Programming expert at http://search390.techtarget.com/ateExperts/ -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
> -Original Message- > From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On > Behalf Of Greg Dyck > Sent: 18 April, 2017 15:58 > To: IBM-MAIN@LISTSERV.UA.EDU > Subject: Re: Paging subsystems in the era of bigass memory > > On 4/18/2017 2:25 AM, Vernooij, Kees - KLM , ITOPT1 wrote: > > As I said, I remember reading this a long time ago, I don't know the > details anymore, whether the source was reliable and whether it is still > working this way. Only a DB2 internal expert should be able to tell. > ... > >> On 12 April 2017 at 10:16, Tom Marchant < > >> 000a2a8c2020-dmarc-requ...@listserv.ua.edu> wrote: > >> [DB2] > >> > >>> So, if it thinks it would be faster to read the record from DASD > than > >> for > >>> MVS to page in the buffer page(s) containing the record, it will > read > >> the > >>> record into different pages that have not been paged out? > > While the DB2 buffer manager code does collect statistics on the > frequency of buffer pages being latched that are currently paged out > (using TPROT), I don't remember seeing any logic to reread a buffer page > vs waiting for a demand page-in to occur in the code. And given that > the current copy of a page may be the one in the buffer pool, rather > than on the copy on DASD, and that multiple threads can have shared > ownership of a buffer pool page, I do not believe this processing exist > (at least for the last 15 years) in DB2. > > 30 years ago this type of logic, to reread from DASD rather than demand > page-in a buffer page, *might* have made sense, when the I/O bandwidth > to the paging subsystem was limited based on the number of page datasets > and a demand page rate of 30-100 pages a second was common to see. But > the world has changed and it would not make sense to do it today. > > Regards, > Greg > Thanks Greg. For information, services and offers, please visit our web site: http://www.klm.com. This e-mail and any attachment may contain confidential and privileged material intended for the addressee only. If you are not the addressee, you are notified that no part of the e-mail or any attachment may be disclosed, copied or distributed, and that any other action related to this e-mail or attachment is strictly prohibited, and may be unlawful. If you have received this e-mail by error, please notify the sender immediately by return e-mail, and delete this message. Koninklijke Luchtvaart Maatschappij NV (KLM), its subsidiaries and/or its employees shall not be liable for the incorrect or incomplete transmission of this e-mail or any attachments, nor responsible for any delay in receipt. Koninklijke Luchtvaart Maatschappij N.V. (also known as KLM Royal Dutch Airlines) is registered in Amstelveen, The Netherlands, with registered number 33014286 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 4/18/2017 2:25 AM, Vernooij, Kees - KLM , ITOPT1 wrote: As I said, I remember reading this a long time ago, I don't know the details anymore, whether the source was reliable and whether it is still working this way. Only a DB2 internal expert should be able to tell. ... On 12 April 2017 at 10:16, Tom Marchant < 000a2a8c2020-dmarc-requ...@listserv.ua.edu> wrote: [DB2] So, if it thinks it would be faster to read the record from DASD than for MVS to page in the buffer page(s) containing the record, it will read the record into different pages that have not been paged out? While the DB2 buffer manager code does collect statistics on the frequency of buffer pages being latched that are currently paged out (using TPROT), I don't remember seeing any logic to reread a buffer page vs waiting for a demand page-in to occur in the code. And given that the current copy of a page may be the one in the buffer pool, rather than on the copy on DASD, and that multiple threads can have shared ownership of a buffer pool page, I do not believe this processing exist (at least for the last 15 years) in DB2. 30 years ago this type of logic, to reread from DASD rather than demand page-in a buffer page, *might* have made sense, when the I/O bandwidth to the paging subsystem was limited based on the number of page datasets and a demand page rate of 30-100 pages a second was common to see. But the world has changed and it would not make sense to do it today. Regards, Greg -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
As I said, I remember reading this a long time ago, I don't know the details anymore, whether the source was reliable and whether it is still working this way. Only a DB2 internal expert should be able to tell. Greg, could you shed a light on this? Kees. > -Original Message- > From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On > Behalf Of Tony Harminc > Sent: 12 April, 2017 18:45 > To: IBM-MAIN@LISTSERV.UA.EDU > Subject: Re: Paging subsystems in the era of bigass memory > > On 12 April 2017 at 10:16, Tom Marchant < > 000a2a8c2020-dmarc-requ...@listserv.ua.edu> wrote: > [DB2] > > > So, if it thinks it would be faster to read the record from DASD than > for > > MVS to page in the buffer page(s) containing the record, it will read > the > > record into different pages that have not been paged out? > > > > It still makes no sense to me. It certainly can't read the record into > the > > same page, because that would require that the page be paged in first. > > > > I've no idea what it *does* do, but it *could* Page Release the old page > before reading from DASD into it. > > > > And what does it do with the old buffer page? Stop using it? Freemain > > and getmain again so that the page slot becomes superfluous? > > > > That would be taken care of by the Page Release. > > Tony H. > > -- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN For information, services and offers, please visit our web site: http://www.klm.com. This e-mail and any attachment may contain confidential and privileged material intended for the addressee only. If you are not the addressee, you are notified that no part of the e-mail or any attachment may be disclosed, copied or distributed, and that any other action related to this e-mail or attachment is strictly prohibited, and may be unlawful. If you have received this e-mail by error, please notify the sender immediately by return e-mail, and delete this message. Koninklijke Luchtvaart Maatschappij NV (KLM), its subsidiaries and/or its employees shall not be liable for the incorrect or incomplete transmission of this e-mail or any attachments, nor responsible for any delay in receipt. Koninklijke Luchtvaart Maatschappij N.V. (also known as KLM Royal Dutch Airlines) is registered in Amstelveen, The Netherlands, with registered number 33014286 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
>I don't want to turn this into a debugging session, but this is what I see in >RMF3 on the non-DB2 utility system. >All of the TPX 'regions' are below this list sorted by Aux Slots. No task >shows any page-ins. Not sure that you haven't already done this: Use IPCS active on that system, and use either ip ilrslotc or panel 2.6i. It'll show you exactly which address space paged out anything to aux (to whatever medium). I recall that in a prior life I could see exactly which asid had a lot of frames on aux (MQ, I think). Since we upgraded to the z13 we increased real drastically on all lpars. Recently, we also started using SCM (flash express). The only thing we were baffled by was the fact that we still needed *one* local page data set to satisfy the system. Other than that, we're very happy with SCM (where we already have it active). Now, to tune sort so that it doesn't need 20 Mod54s as sortwrk anymore when SAS wants to sort 1.600.000.000 records. Barbara -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
I don't want to turn this into a debugging session, but this is what I see in RMF3 on the non-DB2 utility system. All of the TPX 'regions' are below this list sorted by Aux Slots. No task shows any page-ins. -Service-- Frame Occup.-- - Active Frames - AUX PGIN - Jobname C Class Cr TOTAL ACTV IDLE WSET FIXED DIV SLOTS RATE - - *MASTER* S SYSTEM 271K 271K 0 271K 8513 0 526K0 - IOSASS SYSTEM 11702 11702 0 11702 313 0 584360 - ZFS S SYSSTC 10217 10217 0 10217 575 0 351280 - OMCMSS SYSSTC 17310 17310 0 17310 1615 0 314990 - AUTONETV S STCLO 10093 10093 0 10093 452 0 154130 - AUTOAMGR S SYSSTC 2373 2373 0 2373 144 0 125400 - OMMVSS SYSSTC 3968 3968 0 3968 168 0 119290 - TN3270Z S SYSSTC 9993 9993 0 9993 155 0 108010 - RASP S SYSTEM 281 281 0 281 276 0 100680 - OMVS S SYSSTC 4870 4870 0 4870 306 0 89940 - JES2 S SYSSTC 3552 3552 0 3552 3079 0 88810 - RMF S SYSSTC 2884 2884 0 2884 113 0 68270 - NETVIEW S SYSSTC 7620 7620 0 7620 276 0 68150 - TCPIPS SYSSTC 1641 1641 0 1641 141 0 67780 - VLF S SYSSTC 1919 1919 0 1919 104 0 56310 - SMSPDSE S SYSTEM 2852 2852 0 2852 181 0 51800 - RESOLVER S SYSSTC 3209 3209 0 3209 109 0 51360 . . 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 robin...@sce.com -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Graham Harris Sent: Thursday, April 13, 2017 1:41 PM To: IBM-MAIN@LISTSERV.UA.EDU Subject: (External):Re: Paging subsystems in the era of bigass memory Have you got RMF3? Tried looking at STORF to see what address space(s) are consuming all your AUX slots? Are you running Netview? Got a CANZLOG dataspace? That can chew through a lot of memory if not configured quite right in recent z/OS releases. On 13 April 2017 at 20:26, Jesse 1 Robinson mailto:jesse1.robin...@sce.com>> wrote: > This thread did seem to morph into a focus on DB2, but the paging > problem for us is not confined to DB2. We have one utility system that > was set up years ago to be a 'CMC'. It's still dedicated to 'network > stuff', which for some time has narrowed down to CA-TPX, the SNA > session manager. Very little else runs there. Certainly no DB2 or CICS. > Absolutely no end-user apps. > We've sort of ignored this system recently as we turned attention > elsewhere. It was last IPLed in January 2016, well over a year ago! It > runs great except for this burr under the saddle. The local volumes > are all Mod-3. Whatever we decide to do about DB2 will not help here. > > - IEE200I 11.29.28 DISPLAY ASM > - TYPE FULL STAT > - PLPA 100%FULL > - COMMON 36% OK > - LOCAL53% OK > - LOCAL49% OK > - LOCAL43% OK > > . > . > 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 > robin...@sce.com<mailto:robin...@sce.com> > > > -Original Message- > From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] > On Behalf Of Mike Schwab > Sent: Wednesday, April 12, 2017 8:18 PM > To: IBM-MAIN@LISTSERV.UA.EDU<mailto:IBM-MAIN@LISTSERV.UA.EDU> > Subject: (External):Re: Paging subsystems in the era of bigass memory > > Here is an IBM presentation on how to tune z/OS and DB2 memory, > including some parameters to set. > http://www.mdug.org/Presentations/Large%20Memory%20DB2%20Perf%20MDUG.p > df > > On Wed, Apr 12, 2017 at 2:55 PM, Art Gutowski > mailto:arthur.gutow...@gm.com>> > wrote: > > Did someone on this thread say DB2?? > > > > We have been experiencing similar AUX storage creep on our DB2 > > systems, > particularly during LARGE reorgs (more of a gallop than a creep). Our > DB2 guys did some research, opened an ETR with IBM, and found this relic: > > > > Q: > > "[Why was] set realstorage_management to OFF when that zPARM was > introduced in DB2 version 10? > > > > "Details > > "IBM z/OS implemented a Storage Management Design change after DB2 > > v10 > was released. > > "• Before the d
Re: Paging subsystems in the era of bigass memory
On Wed, 12 Apr 2017 22:18:22 -0500, Mike Schwab wrote: >Here is an IBM presentation on how to tune z/OS and DB2 memory, >including some parameters to set. >http://www.mdug.org/Presentations/Large%20Memory%20DB2%20Perf%20MDUG.pdf Thanks for sharing! Regards, Art Gutowski General Motors, LLC -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
Have you got RMF3? Tried looking at STORF to see what address space(s) are consuming all your AUX slots? Are you running Netview? Got a CANZLOG dataspace? That can chew through a lot of memory if not configured quite right in recent z/OS releases. On 13 April 2017 at 20:26, Jesse 1 Robinson wrote: > This thread did seem to morph into a focus on DB2, but the paging problem > for us is not confined to DB2. We have one utility system that was set up > years ago to be a 'CMC'. It's still dedicated to 'network stuff', which for > some time has narrowed down to CA-TPX, the SNA session manager. Very little > else runs there. Certainly no DB2 or CICS. Absolutely no end-user apps. > We've sort of ignored this system recently as we turned attention > elsewhere. It was last IPLed in January 2016, well over a year ago! It runs > great except for this burr under the saddle. The local volumes are all > Mod-3. Whatever we decide to do about DB2 will not help here. > > - IEE200I 11.29.28 DISPLAY ASM > - TYPE FULL STAT > - PLPA 100%FULL > - COMMON 36% OK > - LOCAL53% OK > - LOCAL49% OK > - LOCAL43% OK > > . > . > 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 > robin...@sce.com > > > -Original Message- > From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On > Behalf Of Mike Schwab > Sent: Wednesday, April 12, 2017 8:18 PM > To: IBM-MAIN@LISTSERV.UA.EDU > Subject: (External):Re: Paging subsystems in the era of bigass memory > > Here is an IBM presentation on how to tune z/OS and DB2 memory, including > some parameters to set. > http://www.mdug.org/Presentations/Large%20Memory%20DB2%20Perf%20MDUG.pdf > > On Wed, Apr 12, 2017 at 2:55 PM, Art Gutowski > wrote: > > Did someone on this thread say DB2?? > > > > We have been experiencing similar AUX storage creep on our DB2 systems, > particularly during LARGE reorgs (more of a gallop than a creep). Our DB2 > guys did some research, opened an ETR with IBM, and found this relic: > > > > Q: > > "[Why was] set realstorage_management to OFF when that zPARM was > introduced in DB2 version 10? > > > > "Details > > "IBM z/OS implemented a Storage Management Design change after DB2 v10 > was released. > > "• Before the design change, DB2 used KEEPREAL(NO), virtual storage > pages were really (physically) freed, high CPU cost if YES > > DISCARDDATA KEEPREAL(NO), RSM has to get LPAR level serialization to > > manage those pages that are being freed immediately. That added to CPU > > usage and also caused some CPU spin at the LPAR level to get that > > serialization -- excerpt from PTF > > > > "To get around/minimize the impact of the original design shortcomings > that was introduced by IBM RSM z/OS, setting zPARM realstorage_management > to OFF, would probably have been prudent on most LPARs. HP/EDS tried to > address this new issue IBM created. > > > > "IBM create two PTFs and changed the way DB2 and RSM manages the page > frames. > > > > "• After a design change (now) DB2 uses KEEPREAL(YES), storage is > only virtually freed > > "If DB2 doesn't tell RSM that it doesn't need a frame, then the frame > > will remain backed in real storage in some form. That causes the > > growth of real storage and paging and everything that goes with using > > up REAL storage. KEEPREAL(YES) allows DB2 to tell RSM that z/OS can > > steal the page if it needs it, but DB2 retains ownership of the page, > > and it remains backed with real storage. If z/OS needs the page, it > > can steal it -- excerpt from PTF > > > > "V10 APAR PM88804 APAR PM86862 and PM99575" > > > > So...perhaps check your DSNZPARM and make sure it's coded appropriately > for more modern times. FYI, we are z/OS 2.2 and DB2 11.1, NFM. We are in > the process of rolling out REALSTORAGE_MANAGEMENT=AUTO (the current IBM > recommended setting) across our enterprise. > > > > HTH, > > Art Gutowski (with assist from Doug Drain, Steve Laufer and IBM) > > General Motors, LLC > > > > -- > > For IBM-MAIN subscribe / signoff / archive access instructions, send > > email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN > > > > -- > Mike A Schwab, Springfield IL USA > Where do Forest Rangers go to get away from it all? > > > -- > 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: Paging subsystems in the era of bigass memory
This thread did seem to morph into a focus on DB2, but the paging problem for us is not confined to DB2. We have one utility system that was set up years ago to be a 'CMC'. It's still dedicated to 'network stuff', which for some time has narrowed down to CA-TPX, the SNA session manager. Very little else runs there. Certainly no DB2 or CICS. Absolutely no end-user apps. We've sort of ignored this system recently as we turned attention elsewhere. It was last IPLed in January 2016, well over a year ago! It runs great except for this burr under the saddle. The local volumes are all Mod-3. Whatever we decide to do about DB2 will not help here. - IEE200I 11.29.28 DISPLAY ASM - TYPE FULL STAT - PLPA 100%FULL - COMMON 36% OK - LOCAL53% OK - LOCAL49% OK - LOCAL43% OK . . 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 robin...@sce.com -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Mike Schwab Sent: Wednesday, April 12, 2017 8:18 PM To: IBM-MAIN@LISTSERV.UA.EDU Subject: (External):Re: Paging subsystems in the era of bigass memory Here is an IBM presentation on how to tune z/OS and DB2 memory, including some parameters to set. http://www.mdug.org/Presentations/Large%20Memory%20DB2%20Perf%20MDUG.pdf On Wed, Apr 12, 2017 at 2:55 PM, Art Gutowski wrote: > Did someone on this thread say DB2?? > > We have been experiencing similar AUX storage creep on our DB2 systems, > particularly during LARGE reorgs (more of a gallop than a creep). Our DB2 > guys did some research, opened an ETR with IBM, and found this relic: > > Q: > "[Why was] set realstorage_management to OFF when that zPARM was introduced > in DB2 version 10? > > "Details > "IBM z/OS implemented a Storage Management Design change after DB2 v10 was > released. > "• Before the design change, DB2 used KEEPREAL(NO), virtual storage > pages were really (physically) freed, high CPU cost if YES > DISCARDDATA KEEPREAL(NO), RSM has to get LPAR level serialization to > manage those pages that are being freed immediately. That added to CPU > usage and also caused some CPU spin at the LPAR level to get that > serialization -- excerpt from PTF > > "To get around/minimize the impact of the original design shortcomings that > was introduced by IBM RSM z/OS, setting zPARM realstorage_management to OFF, > would probably have been prudent on most LPARs. HP/EDS tried to address this > new issue IBM created. > > "IBM create two PTFs and changed the way DB2 and RSM manages the page frames. > > "• After a design change (now) DB2 uses KEEPREAL(YES), storage is only > virtually freed > "If DB2 doesn't tell RSM that it doesn't need a frame, then the frame > will remain backed in real storage in some form. That causes the > growth of real storage and paging and everything that goes with using > up REAL storage. KEEPREAL(YES) allows DB2 to tell RSM that z/OS can > steal the page if it needs it, but DB2 retains ownership of the page, > and it remains backed with real storage. If z/OS needs the page, it > can steal it -- excerpt from PTF > > "V10 APAR PM88804 APAR PM86862 and PM99575" > > So...perhaps check your DSNZPARM and make sure it's coded appropriately for > more modern times. FYI, we are z/OS 2.2 and DB2 11.1, NFM. We are in the > process of rolling out REALSTORAGE_MANAGEMENT=AUTO (the current IBM > recommended setting) across our enterprise. > > HTH, > Art Gutowski (with assist from Doug Drain, Steve Laufer and IBM) > General Motors, LLC > > -- > For IBM-MAIN subscribe / signoff / archive access instructions, send > email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN -- Mike A Schwab, Springfield IL USA Where do Forest Rangers go to get away from it all? -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
Just check with our DB2 SYSPROG, same here. Carmen - Original Message - From: "Mike Schwab" To: IBM-MAIN@LISTSERV.UA.EDU Sent: Wednesday, April 12, 2017 10:18:22 PM Subject: Re: Paging subsystems in the era of bigass memory Here is an IBM presentation on how to tune z/OS and DB2 memory, including some parameters to set. http://www.mdug.org/Presentations/Large%20Memory%20DB2%20Perf%20MDUG.pdf On Wed, Apr 12, 2017 at 2:55 PM, Art Gutowski wrote: > Did someone on this thread say DB2?? > > We have been experiencing similar AUX storage creep on our DB2 systems, > particularly during LARGE reorgs (more of a gallop than a creep). Our DB2 > guys did some research, opened an ETR with IBM, and found this relic: > > Q: > "[Why was] set realstorage_management to OFF when that zPARM was introduced > in DB2 version 10? > > "Details > "IBM z/OS implemented a Storage Management Design change after DB2 v10 was > released. > "• Before the design change, DB2 used KEEPREAL(NO), virtual storage pages > were really (physically) freed, high CPU cost if YES > DISCARDDATA KEEPREAL(NO), RSM has to get LPAR level serialization to manage > those pages that are being freed immediately. That added to CPU usage and > also caused some CPU spin at the LPAR level to get that serialization -- > excerpt from PTF > > "To get around/minimize the impact of the original design shortcomings that > was introduced by IBM RSM z/OS, setting zPARM realstorage_management to OFF, > would probably have been prudent on most LPARs. HP/EDS tried to address this > new issue IBM created. > > "IBM create two PTFs and changed the way DB2 and RSM manages the page frames. > > "• After a design change (now) DB2 uses KEEPREAL(YES), storage is only > virtually freed > "If DB2 doesn't tell RSM that it doesn't need a frame, then the frame will > remain backed in real storage in some form. That causes the growth of real > storage and paging and everything that goes with using up REAL storage. > KEEPREAL(YES) allows DB2 to tell RSM that z/OS can steal the page if it needs > it, but DB2 retains ownership of the page, and it remains backed with real > storage. If z/OS needs the page, it can steal it -- excerpt from PTF > > "V10 APAR PM88804 APAR PM86862 and PM99575" > > So...perhaps check your DSNZPARM and make sure it's coded appropriately for > more modern times. FYI, we are z/OS 2.2 and DB2 11.1, NFM. We are in the > process of rolling out REALSTORAGE_MANAGEMENT=AUTO (the current IBM > recommended setting) across our enterprise. > > HTH, > Art Gutowski (with assist from Doug Drain, Steve Laufer and IBM) > General Motors, LLC > > -- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN -- Mike A Schwab, Springfield IL USA Where do Forest Rangers go to get away from it all? -- 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: Paging subsystems in the era of bigass memory
Here is an IBM presentation on how to tune z/OS and DB2 memory, including some parameters to set. http://www.mdug.org/Presentations/Large%20Memory%20DB2%20Perf%20MDUG.pdf On Wed, Apr 12, 2017 at 2:55 PM, Art Gutowski wrote: > Did someone on this thread say DB2?? > > We have been experiencing similar AUX storage creep on our DB2 systems, > particularly during LARGE reorgs (more of a gallop than a creep). Our DB2 > guys did some research, opened an ETR with IBM, and found this relic: > > Q: > "[Why was] set realstorage_management to OFF when that zPARM was introduced > in DB2 version 10? > > "Details > "IBM z/OS implemented a Storage Management Design change after DB2 v10 was > released. > "• Before the design change, DB2 used KEEPREAL(NO), virtual storage > pages were really (physically) freed, high CPU cost if YES > DISCARDDATA KEEPREAL(NO), RSM has to get LPAR level serialization to manage > those pages that are being freed immediately. That added to CPU usage and > also caused some CPU spin at the LPAR level to get that serialization -- > excerpt from PTF > > "To get around/minimize the impact of the original design shortcomings that > was introduced by IBM RSM z/OS, setting zPARM realstorage_management to OFF, > would probably have been prudent on most LPARs. HP/EDS tried to address this > new issue IBM created. > > "IBM create two PTFs and changed the way DB2 and RSM manages the page frames. > > "• After a design change (now) DB2 uses KEEPREAL(YES), storage is only > virtually freed > "If DB2 doesn't tell RSM that it doesn't need a frame, then the frame will > remain backed in real storage in some form. That causes the growth of real > storage and paging and everything that goes with using up REAL storage. > KEEPREAL(YES) allows DB2 to tell RSM that z/OS can steal the page if it needs > it, but DB2 retains ownership of the page, and it remains backed with real > storage. If z/OS needs the page, it can steal it -- excerpt from PTF > > "V10 APAR PM88804 APAR PM86862 and PM99575" > > So...perhaps check your DSNZPARM and make sure it's coded appropriately for > more modern times. FYI, we are z/OS 2.2 and DB2 11.1, NFM. We are in the > process of rolling out REALSTORAGE_MANAGEMENT=AUTO (the current IBM > recommended setting) across our enterprise. > > HTH, > Art Gutowski (with assist from Doug Drain, Steve Laufer and IBM) > General Motors, LLC > > -- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN -- Mike A Schwab, Springfield IL USA Where do Forest Rangers go to get away from it all? -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: EXTERNAL: Re: Paging subsystems in the era of bigass memory
Ditto ! Thanks Art Jerry Whitteridge Manager Mainframe Systems & Storage Albertsons - Safeway Inc. 623 869 5523 Corporate Tieline - 85523 If you feel in control you just aren't going fast enough. -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Carmen Vitullo Sent: Wednesday, April 12, 2017 1:02 PM To: IBM-MAIN@LISTSERV.UA.EDU Subject: EXTERNAL: Re: Paging subsystems in the era of bigass memory Yes, and Thank you Art for that - I've passed this info on to our DB2 SYSPROG Carmen - Original Message - From: "Art Gutowski" To: IBM-MAIN@LISTSERV.UA.EDU Sent: Wednesday, April 12, 2017 2:55:53 PM Subject: Re: Paging subsystems in the era of bigass memory Did someone on this thread say DB2?? We have been experiencing similar AUX storage creep on our DB2 systems, particularly during LARGE reorgs (more of a gallop than a creep). Our DB2 guys did some research, opened an ETR with IBM, and found this relic: Q: "[Why was] set realstorage_management to OFF when that zPARM was introduced in DB2 version 10? "Details "IBM z/OS implemented a Storage Management Design change after DB2 v10 was released. "• Before the design change, DB2 used KEEPREAL(NO), virtual storage pages were really (physically) freed, high CPU cost if YES DISCARDDATA KEEPREAL(NO), RSM has to get LPAR level serialization to manage those pages that are being freed immediately. That added to CPU usage and also caused some CPU spin at the LPAR level to get that serialization -- excerpt from PTF "To get around/minimize the impact of the original design shortcomings that was introduced by IBM RSM z/OS, setting zPARM realstorage_management to OFF, would probably have been prudent on most LPARs. HP/EDS tried to address this new issue IBM created. "IBM create two PTFs and changed the way DB2 and RSM manages the page frames. "• After a design change (now) DB2 uses KEEPREAL(YES), storage is only virtually freed "If DB2 doesn't tell RSM that it doesn't need a frame, then the frame will remain backed in real storage in some form. That causes the growth of real storage and paging and everything that goes with using up REAL storage. KEEPREAL(YES) allows DB2 to tell RSM that z/OS can steal the page if it needs it, but DB2 retains ownership of the page, and it remains backed with real storage. If z/OS needs the page, it can steal it -- excerpt from PTF "V10 APAR PM88804 APAR PM86862 and PM99575" So...perhaps check your DSNZPARM and make sure it's coded appropriately for more modern times. FYI, we are z/OS 2.2 and DB2 11.1, NFM. We are in the process of rolling out REALSTORAGE_MANAGEMENT=AUTO (the current IBM recommended setting) across our enterprise. HTH, Art Gutowski (with assist from Doug Drain, Steve Laufer and IBM) General Motors, LLC -- 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 Warning: All e-mail sent to this address will be received by the corporate e-mail system, and is subject to archival and review by someone other than the recipient. This e-mail may contain proprietary information and is intended only for the use of the intended recipient(s). If the reader of this message is not the intended recipient(s), you are notified that you have received this message in error and that any review, dissemination, distribution or copying of this message is strictly prohibited. If you have received this message in error, please notify the sender immediately. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
Yes, and Thank you Art for that - I've passed this info on to our DB2 SYSPROG Carmen - Original Message - From: "Art Gutowski" To: IBM-MAIN@LISTSERV.UA.EDU Sent: Wednesday, April 12, 2017 2:55:53 PM Subject: Re: Paging subsystems in the era of bigass memory Did someone on this thread say DB2?? We have been experiencing similar AUX storage creep on our DB2 systems, particularly during LARGE reorgs (more of a gallop than a creep). Our DB2 guys did some research, opened an ETR with IBM, and found this relic: Q: "[Why was] set realstorage_management to OFF when that zPARM was introduced in DB2 version 10? "Details "IBM z/OS implemented a Storage Management Design change after DB2 v10 was released. "• Before the design change, DB2 used KEEPREAL(NO), virtual storage pages were really (physically) freed, high CPU cost if YES DISCARDDATA KEEPREAL(NO), RSM has to get LPAR level serialization to manage those pages that are being freed immediately. That added to CPU usage and also caused some CPU spin at the LPAR level to get that serialization -- excerpt from PTF "To get around/minimize the impact of the original design shortcomings that was introduced by IBM RSM z/OS, setting zPARM realstorage_management to OFF, would probably have been prudent on most LPARs. HP/EDS tried to address this new issue IBM created. "IBM create two PTFs and changed the way DB2 and RSM manages the page frames. "• After a design change (now) DB2 uses KEEPREAL(YES), storage is only virtually freed "If DB2 doesn't tell RSM that it doesn't need a frame, then the frame will remain backed in real storage in some form. That causes the growth of real storage and paging and everything that goes with using up REAL storage. KEEPREAL(YES) allows DB2 to tell RSM that z/OS can steal the page if it needs it, but DB2 retains ownership of the page, and it remains backed with real storage. If z/OS needs the page, it can steal it -- excerpt from PTF "V10 APAR PM88804 APAR PM86862 and PM99575" So...perhaps check your DSNZPARM and make sure it's coded appropriately for more modern times. FYI, we are z/OS 2.2 and DB2 11.1, NFM. We are in the process of rolling out REALSTORAGE_MANAGEMENT=AUTO (the current IBM recommended setting) across our enterprise. HTH, Art Gutowski (with assist from Doug Drain, Steve Laufer and IBM) General Motors, LLC -- 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: Paging subsystems in the era of bigass memory
Did someone on this thread say DB2?? We have been experiencing similar AUX storage creep on our DB2 systems, particularly during LARGE reorgs (more of a gallop than a creep). Our DB2 guys did some research, opened an ETR with IBM, and found this relic: Q: "[Why was] set realstorage_management to OFF when that zPARM was introduced in DB2 version 10? "Details "IBM z/OS implemented a Storage Management Design change after DB2 v10 was released. "• Before the design change, DB2 used KEEPREAL(NO), virtual storage pages were really (physically) freed, high CPU cost if YES DISCARDDATA KEEPREAL(NO), RSM has to get LPAR level serialization to manage those pages that are being freed immediately. That added to CPU usage and also caused some CPU spin at the LPAR level to get that serialization -- excerpt from PTF "To get around/minimize the impact of the original design shortcomings that was introduced by IBM RSM z/OS, setting zPARM realstorage_management to OFF, would probably have been prudent on most LPARs. HP/EDS tried to address this new issue IBM created. "IBM create two PTFs and changed the way DB2 and RSM manages the page frames. "• After a design change (now) DB2 uses KEEPREAL(YES), storage is only virtually freed "If DB2 doesn't tell RSM that it doesn't need a frame, then the frame will remain backed in real storage in some form. That causes the growth of real storage and paging and everything that goes with using up REAL storage. KEEPREAL(YES) allows DB2 to tell RSM that z/OS can steal the page if it needs it, but DB2 retains ownership of the page, and it remains backed with real storage. If z/OS needs the page, it can steal it -- excerpt from PTF "V10 APAR PM88804 APAR PM86862 and PM99575" So...perhaps check your DSNZPARM and make sure it's coded appropriately for more modern times. FYI, we are z/OS 2.2 and DB2 11.1, NFM. We are in the process of rolling out REALSTORAGE_MANAGEMENT=AUTO (the current IBM recommended setting) across our enterprise. HTH, Art Gutowski (with assist from Doug Drain, Steve Laufer and IBM) General Motors, LLC -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
The topic seems to of morphed from Do I need to acheive a 30% page dataset utilization from virtual storage stable LPAR with a single large DB2 to Running many DB2s in a single LPAR I seem to have a memory leak that requires periodic IPLs to avoid an aux storage problem other than some common words like LPAR and DB2 the issues are not the same. Avram Friedman -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
For you, I hope that never happens, I have some AUX pages that never free, and creep 1% or so every 2 days, been following the discussion and trying to undersand why suddenly DB2 seems to be the culprit, when I've seen DB2 workloads not have an adverse effect on AUX storage, my last IPL of this system was Feb 26th, and we have a Large Frame area defined 7740M my locals are on Mod9's LOCAL 20% OK 4371 PAGE.SYSA.VSYSPG1.LOCAL LOCAL 20% OK 4471 PAGE.SYSA.VSYSPG2.LOCAL LOCAL 20% OK 4571 PAGE.SYSA.VSYSPG3.LOCAL LOCAL 19% OK 4671 PAGE.SYSA.VSYSPG4.LOCAL LOCAL 19% OK 4771 PAGE.SYSA.VSYSPG5.LOCAL LOCAL 19% OK 4072 PAGE.SYSA.VSYSPG6.LOCAL LOCAL 20% OK 4172 PAGE.SYSA.VSYSPG7.LOCAL LOCAL 20% OK 4272 PAGE.SYSA.VSYSPG8.LOCAL LOCAL 20% OK 4070 PAGE.SYSA.VSYSPGH.LOCAL LOCAL 20% OK 4170 PAGE.SYSA.VSYSPGI.LOCAL LOCAL 20% OK 4270 PAGE.SYSA.VSYSPGJ.LOCAL LOCAL 20% OK 4370 PAGE.SYSA.VSYSPGK.LOCAL LOCAL 20% OK 4470 PAGE.SYSA.VSYSPGL.LOCAL LOCAL 20% OK 4570 PAGE.SYSA.VSYSPGM.LOCAL we have 8 DB2MSTR regions 43 misc regions total (DISTR,IMS...) if I decide not to IPL until June or July I see myself running the risk of getting AUX storage shortages what to do? add some more locals? NO! Carmen - Original Message - From: "Jesse 1 Robinson" To: IBM-MAIN@LISTSERV.UA.EDU Sent: Wednesday, April 12, 2017 12:56:19 PM Subject: Re: Paging subsystems in the era of bigass memory I'd like to pull this discussion back up to the jet stream level. We face the prospect of having to IPL z/OS just to relieve an ASM shortage. Weenie-ware servers seem to run forever without this encumbrance. We put our highest-value mission-critical applications on a platform that cannot keep its own shorts tidy. Enough already. Even if IBM gave us memory for free, we would still have endure interruptions to get it installed. And how much is sufficient? How many interruptions? Throwing more hardware at a software deficiency is not a solution. I'm afraid that some MBA PFK will propose running the whole shebang on *nix. Then where will we (all) end up? . . 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 robin...@sce.com -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Tom Marchant Sent: Wednesday, April 12, 2017 10:15 AM To: IBM-MAIN@LISTSERV.UA.EDU Subject: (External):Re: Paging subsystems in the era of bigass memory On Wed, 12 Apr 2017 12:44:32 -0400, Tony Harminc wrote: >On 12 April 2017 at 10:16, Tom Marchant < >000a2a8c2020-dmarc-requ...@listserv.ua.edu> wrote: >[DB2] > >> It still makes no sense to me. It certainly can't read the record >> into the same page, because that would require that the page be paged in >> first. > >I've no idea what it *does* do, but it *could* Page Release the old >page before reading from DASD into it. Thanks, Tony. And BTW, Kees, I should have said "I don't understand" rather than "It makes no sense to me." Subtle difference, perhaps. -- Tom Marchant -- 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: Paging subsystems in the era of bigass memory
I'd like to pull this discussion back up to the jet stream level. We face the prospect of having to IPL z/OS just to relieve an ASM shortage. Weenie-ware servers seem to run forever without this encumbrance. We put our highest-value mission-critical applications on a platform that cannot keep its own shorts tidy. Enough already. Even if IBM gave us memory for free, we would still have endure interruptions to get it installed. And how much is sufficient? How many interruptions? Throwing more hardware at a software deficiency is not a solution. I'm afraid that some MBA PFK will propose running the whole shebang on *nix. Then where will we (all) end up? . . 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 robin...@sce.com -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Tom Marchant Sent: Wednesday, April 12, 2017 10:15 AM To: IBM-MAIN@LISTSERV.UA.EDU Subject: (External):Re: Paging subsystems in the era of bigass memory On Wed, 12 Apr 2017 12:44:32 -0400, Tony Harminc wrote: >On 12 April 2017 at 10:16, Tom Marchant < >000a2a8c2020-dmarc-requ...@listserv.ua.edu> wrote: >[DB2] > >> It still makes no sense to me. It certainly can't read the record >> into the same page, because that would require that the page be paged in >> first. > >I've no idea what it *does* do, but it *could* Page Release the old >page before reading from DASD into it. Thanks, Tony. And BTW, Kees, I should have said "I don't understand" rather than "It makes no sense to me." Subtle difference, perhaps. -- Tom Marchant -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
000a2a8c2020-dmarc-requ...@listserv.ua.edu (Tom Marchant) writes: > Are you suggesting that before DB2 references a page containing a > buffer, it checks to see if it is paged out? And that if it is paged out, > it doesn't use the record in the buffer, but instead reads it into a > different page? That makes no sense to me. there is separate problem that i've repeatedly pontificated about since the 70s ... originally involving running MVS in VM370 virtual machine. If VM370 is running a "LRU" (least recently used) replacement algorithm and MVS (in virtual machine) is running a "LRU" (least recently used) replacement algorithm ... then MVS will tend to select to replace and use a virtual machine page that hasn't been used in a long time ... which is also the page that vm370 will have chosen to have replaced and paged-out. LRU recently used algorithm assumes that a page that hasn't been used for a long time is the most likely not be used for a long time in the future. However, if a LRU is running in memory space managed by a LRU algorithm ... then the 2nd level LRU is likely to choose the next page to use ... the exact same page that the lower level LRU has removed from storage a virtual LRU algorithm starts to look like MRU algorithm, aka the least recently used page becomes the page that is most likely to be used next (as opposed to the least likely to be used next). past posts mentioning page replacement algorithm http://www.garlic.com/~lynn/subtopic.html#clock I was also involved in the original sql/relational implementation, System/R ... some past posts http://www.garlic.com/~lynn/submain.html#systemr which was done on 370/145 running under VM370. The issue was that the RDBMS (system/r) running in virtual address space manages its own cache of records with algorithms similar to LRU ... which is running in virtual memory managed by VM370 with an LRU algorithm. Then the next least recently used cache area that RDBMS is likely to use ... is also the least recently used area that VM370 has likely to have been replaced & removed from real memory (a LRU managed area running in a virtual memory LRU managed area, will violate assumption that the least recently used area is the least likely to be used in the future and can be replaced). trivia, managed to do System/R technology transfer to Endicott for SQL/DS "under the radar", while the corporation was pre-occupied doing the next official follow-on DBMS "EAGLE". Then when "EAGLE" effort imploded there was request for how fast could SystemR/SQLDS be ported to MVS ... which is then later announced as DB2 (originally for decision support *ONLY*). DB2 will have its own area for caching DBMS records ... managed with a LRU-like algorithm ... running in an MVS virtual memory managed with an LRU-like algorithm ... assuming the least recently used DB2 cache area is then the least likely to be used in the future and can be removed from processor memory. When Jim Gray leaves for Tandem ... he palms off some number of things on to me ... included DBMS consulting with the IMS group ... old email http://www.garlic.com/~lynn/2007.html#email801016 and recent long-winded post (facebook IBM discussion) about Jim Gray (after he left for Tandem) and page replacement/management algorithms http://www.garlic.com/~lynn/2017d.html#52 Some IBM Research RJ reports At DEC81 ACM SIGOPS, Jim asks if I could help a tandem co-worker get his stanford PHD that involves quite a bit of "global LRU" ... and some "local LRU" forces were pressuring his thesis advisor to not approve the work. I had done a lot of work on page replacement as undergraduate in the 60s ... including "global LRU" about the same time there were academic papers on "local LRU" were being published in ACM Communications ... and I had direct apples-to-apples comparisons between "global" and "local" ... showing "global LRU" was significantly better than "local LRU". Unfortunately IBM research management prevented me from responded for nearly a year (even tho the work had been done before joining IBM) ... possibly thinking they were punishing me (which would be better than they were doing it because they were taking sides in the academic dispute). finally was able to send response the following Oct. http://www.garlic.com/~lynn/2006w.html#email821019 trivia: I was being blamed for online computer conferencing (precursor to discussion mailing lists like ibm-main as well as social media) on the IBM internal network (larger than arpanet/internet from just about the beginning until sometime mid-80s) in the late 70s and early 80s ... folklore is that when IBM corporate executive committee was told about online computer conferencing (and the internal network), 5of6 wanted to fire me. i.e. IBM management may have considered that blocking my response for nearly a year was part of punishment for doing online computer conferencing. some past posts http://www.garlic.com/~lynn/subnetwork.html#cmc not getting fired for
Re: Paging subsystems in the era of bigass memory
On Wed, 12 Apr 2017 12:44:32 -0400, Tony Harminc wrote: >On 12 April 2017 at 10:16, Tom Marchant < >000a2a8c2020-dmarc-requ...@listserv.ua.edu> wrote: >[DB2] > >> It still makes no sense to me. It certainly can't read the record into the >> same page, because that would require that the page be paged in first. > >I've no idea what it *does* do, but it *could* Page Release the old page >before reading from DASD into it. Thanks, Tony. And BTW, Kees, I should have said "I don't understand" rather than "It makes no sense to me." Subtle difference, perhaps. -- Tom Marchant -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 12 April 2017 at 10:16, Tom Marchant < 000a2a8c2020-dmarc-requ...@listserv.ua.edu> wrote: [DB2] > So, if it thinks it would be faster to read the record from DASD than for > MVS to page in the buffer page(s) containing the record, it will read the > record into different pages that have not been paged out? > > It still makes no sense to me. It certainly can't read the record into the > same page, because that would require that the page be paged in first. > I've no idea what it *does* do, but it *could* Page Release the old page before reading from DASD into it. > And what does it do with the old buffer page? Stop using it? Freemain > and getmain again so that the page slot becomes superfluous? > That would be taken care of by the Page Release. Tony H. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 04/12/2017 10:16 AM, Paul Gilmartin wrote: > On Wed, 12 Apr 2017 09:50:44 -0500, Joel C. Ewing wrote: >> GETMAIN returns a success indication based purely on whether virtual >> storage is available to the address space within the REGION size >> constrains. >> > Does it not check for available paging space? Does this imply that your > job and my job can each do a GETMAIN for storage which in the aggregate > exceeds the available paging space and both will succeed? > >> It is the SysProg's non-trivial job to configure the LPAR real memory >> and configure z/OS to adequately constrain max REGION sizes, the number >> of address spaces and size the paging subsystem so that page thrashing >> doesn't occur and critical page slot thresholds in the paging subsystem >> aren't exceeded. There are tools, configuration parameters, and exits >> that make that possible. >> > Then if you job touches all its pages then my job attempts to touch all > its pages, may I experience a failure because there are no slots to page > something out? > > -- gil To reach a scenario where two jobs could potentially interfere with each other in that way would seemingly indicate a serious failure in configuration if two ill-behaved jobs with such large REGION constraints were allowed to run concurrently without an adequately sized page data set pool to support them. And the implication that each would also impose a serious impact on real storage by them self further suggests that such jobs should be constrained to a job class that disallows two running concurrently. Any job or subsystem deliberately designed to use large amounts of virtual memory above the bar or a significant percentage of the real memory assigned to an LPAR should be under the control of people smart enough to know they need to coordinate their plans to the SysProgs that support that LPAR so that things can be sized appropriately and new page datasets can be created and added. If one is in an environment where you can't trust that communication to take place, then the SysProgs need to add REGION size enforcement so that people are forced to communicate before their jobs can get authority to successfully allocate enough virtual storage to become a problem. I don't think your job would fail immediately, just wait. I believe there would be all sorts of nasty console messages which start at a fairly low level of page slots in use (30%?) and maybe even include indications if one job consumes an unreasonable percentage of aux paging pool space, so there is potentially some advance warning. If the operator were able to dynamically add a sufficient number of new page data sets, or cancel one of the jobs causing the problem, things should be able to proceed. If the page auxiliary storage becomes 100% full and there is no way to add more space in time or determine the offending address spaces and cancel them, I suspect the whole system, not just your job, is toast. In the environment I worked, the only really big users of virtual storage at the time were CICS and DB2, and the people supporting those subsystems understood quite well the need to communicate to z/OS support any significant plans to change their virtual/real storage usage. If other application areas are planning to use 64-bit virtual storage, they also need to be properly educated to communicate their plans and understand the consequences of failing to do so. Joel C. Ewing -- Joel C. Ewing,Bentonville, AR jcew...@acm.org -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
ibmsysp...@ibm-sys-prog.com (Avram Friedman) writes: > While they do not grow in in perfect lock step The presence of Big ass > memory comes with big ass dasd volumes (these are the technical terms > of course) Do you know 3350's and 2314s were once used as paging > devices? For that matter do you no the number 2314 was chosen as the > # for a model of disk drive because it was 4 times larger than its > asexual parent the 2311. original CP67 paged on 2301 fixed-head/track drum ... but it was only 4mbytes ... and "overflow" to 2314. The original code did single page transfer per sio and pure fifo. as undergraduate in the 60s ... i redid 2301 support for "chained requests", so it ordered queued request for maximum transfers per revolution and multiple transfer per sio (both 2301 & 2314 for same arm position) and added ordered seek queueing for 2314. 2301 peak transfers went from about 80/sec to nearly media transfer ... around 270/sec. 2314 thruput increased 2-3 times and degradation (service time increase as load increased) was much more graceful. trivia: later, 3350 had small fixed-head/track option/feature that could be used for paging. however, it didn't have "multiple exposure" like 2305 (fixed-head/track disk, basically multiple subchannel addresses for same device, that could do things like transfers while other requests were still waiting for rotation). I had proposal to do 3350 multiple exposures that would allow doing data transfers for fixed-head area overlapped with arm seek motion. http://www-03.ibm.com/ibm/history/exhibits/storage/storage_3350.html however, there was group in POK that was planning on doing "vulcan" ... an "electronic disk" ... on the lines of "1655" mentioned upthread http://www.garlic.com/~lynn/2017d.html#63 and blocked the 3350 multiple-exposure feature. They then got killed when they were told that customers were ordering all the memory that IBM could produce for processor memory (at higher markup) ... but I wasn't allowed to resurrect the multiple-exposure. As it happened, the "1655" vendor was using memory chips that failed processor memory tests, but could be still be used for electronic disk. longer discussion http://www.garilc.com/~lynn/2006s.html#45 more 2301 mention in this recent comp.arch post (2301 was sort of 2303 but read/wrote four tracks in parallel, 4 times the data transfer rate, 1/4th the number of tracks, each track four times larger) http://www.garlic.com/~lynn/2017d.html#60 Optimizing the Hard Disk Directly later 2305-1 did something similar ... it put every other head on the same track but offset 180degrees ... resulting in half the number of tracks (and half the total capacity) ... and alternatating bytes were on opposides sides. Avg. rotation delay was then only quarter of track (rather than half) ... and data transfer was twice (3mbytes/sec rather than 1.5mbytes/sec) ... doing read/writes from both heads in parallel (on opposite sides of the track, odd on one side, even on the other side). http://www-03.ibm.com/ibm/history/exhibits/storage/storage_2305.html -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On Wed, 12 Apr 2017 09:50:44 -0500, Joel C. Ewing wrote: > >GETMAIN returns a success indication based purely on whether virtual >storage is available to the address space within the REGION size >constrains. > Does it not check for available paging space? Does this imply that your job and my job can each do a GETMAIN for storage which in the aggregate exceeds the available paging space and both will succeed? >It is the SysProg's non-trivial job to configure the LPAR real memory >and configure z/OS to adequately constrain max REGION sizes, the number >of address spaces and size the paging subsystem so that page thrashing >doesn't occur and critical page slot thresholds in the paging subsystem >aren't exceeded. There are tools, configuration parameters, and exits >that make that possible. > Then if you job touches all its pages then my job attempts to touch all its pages, may I experience a failure because there are no slots to page something out? -- gil -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 04/11/2017 05:41 PM, Paul Gilmartin wrote: > On Tue, 11 Apr 2017 16:45:10 -0500, Greg Dyck wrote: > >> On 4/11/2017 3:26 PM, Paul Gilmartin wrote: >>> My understanding, ancient, probably outdated, and certainly naive is that >>> there is little communication between GETMAIN/FREEMAIN and the paging >>> subsystem. If a program touches a page that was never GETMAINed no error >>> occurs; simply a page slot is allocated up to the limit of the REGION >>> parameter. Conversely, on FREEMAIN the page slots are not released. >> *Totally* wrong. z/OS (or any predecessors or relatives like VS1) has >> never worked like you describe. >> >> If a virtual page is not 'getmain assigned' by VSM it will never be >> backed by a real frame by RSM and an 0C4-11 will occur. When all >> virtual storage on a page is freed VSM will return the allocated AUX >> slot and real frame, if either have been allocated, and reset the >> 'getmain assigned' indicator. >> > What about the other side? Will GETMAIN return indication of success > only if the requested page slots can be committed? > > A Google search for "lazy malloc" shows the preponderance of complaints > concern Linux. > > My information came from a UNIX/C programmer before the ascendancy > of Linux. > > And from an MVS wizard who insisted that for GETMAIN to fail if requested > page slots could not be committed would cause enormous breakage of > existing art. > > -- gil > > How would GETMAIN have any way of predicting either the future availability of page frames when a virtual storage page is later referenced or the future availability of paging system page slots if and when z/OS decides to page out a backing page frame? I seem to recall there are some inhibitions on starting new address spaces in a "storage constrained" environment, but that doesn't apply to individual GETMAINs from an already running address spaces. GETMAIN returns a success indication based purely on whether virtual storage is available to the address space within the REGION size constrains. It is the SysProg's non-trivial job to configure the LPAR real memory and configure z/OS to adequately constrain max REGION sizes, the number of address spaces and size the paging subsystem so that page thrashing doesn't occur and critical page slot thresholds in the paging subsystem aren't exceeded. There are tools, configuration parameters, and exits that make that possible. Joel C. Ewing -- Joel C. Ewing,Bentonville, AR jcew...@acm.org -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On Wed, 12 Apr 2017 13:08:32 +, Vernooij, Kees wrote: > >> -Original Message- >> From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On >> Behalf Of Tom Marchant >> Sent: 12 April, 2017 15:03 >> To: IBM-MAIN@LISTSERV.UA.EDU >> Subject: Re: Paging subsystems in the era of bigass memory >> >> On Wed, 12 Apr 2017 06:28:05 +, Vernooij, Kees wrote: >> >> >From here, there the story still goes on IIRC: if DB2 again needs >> >the data, it would be paged in in any normal task. However, DB2 >> >is more intelligent: it keeps track of how long it takes to page-in >> >the page or read it again from disk, in a I/O that is already running. >> >If the latter is faster this is used and the aux slot is really >> useless. >> >> Are you suggesting that before DB2 references a page containing a >> buffer, it checks to see if it is paged out? > >Yes. > >> And that if it is paged >> out, >> it doesn't use the record in the buffer, but instead reads it into a >> different page? > >No, it checks its(?) statistics about page-in time and reading it directly >from dasd and then decides which will be faster. I heard this years ago, >when heavy paging systems might produce slow page-ins and adding a >page to an in-progress pre-fetch could be faster. So, if it thinks it would be faster to read the record from DASD than for MVS to page in the buffer page(s) containing the record, it will read the record into different pages that have not been paged out? It still makes no sense to me. It certainly can't read the record into the same page, because that would require that the page be paged in first. And what does it do with the old buffer page? Stop using it? Freemain and getmain again so that the page slot becomes superfluous? > >> That makes no sense to me. > >This does? What? -- Tom Marchant -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
While they do not grow in in perfect lock step The presence of Big ass memory comes with big ass dasd volumes (these are the technical terms of course) Do you know 3350's and 2314s were once used as paging devices? For that matter do you no the number 2314 was chosen as the # for a model of disk drive because it was 4 times larger than its asexual parent the 2311. You may be aware that the reason for new addressing architecture like 32 and 64 bit with corresponding changes to memory mapping came as a result of addressing older storage constraints. Typically it was CICS, IMS and DB2 that were contained and the newer architectures are designed to hold, you guessed it CICS, IMS, and DB2. The size of spinning page datasets is related to the amount of virutual memory in use. Have you noticed that the use of the LPA dataset in any shop does not change after IPL? That is because LPA is built at IPL time and can even be skipped by not doing a CLPA. LPA does not require extra space for minute to minute operations because it does not change. It may need space if a software update is done. You mention that you have a near zero paging rate. With that near zero rate for locals those datasets can enjoy some of the same allocation freedoms the LPA have. If it is not going to grow one does not have to add much for minute to minute growth. Most old rules of thumb are based on a TSO environment with lots of swapable address spaces. This is not the case for large DBMS and TP monitors. Avram Friedman On Tue, 11 Apr 2017 10:46:40 -0400, Pinnacle wrote: >Gone are the halcyon days when we could run an LPAR with three mod-3's >as the local paging subsystem. With today's large memory sizes, I'm >faced with having to completely rethink my paging subsystems. I've >currently got a 133GB LPAR with 18 mod-9 locals at 44%. I'm going to >add 22 more mod-9's, which will get me just under the 30% threshold. >That's 40 page datasets, which is about 30 more than the most I've ever >managed. I'm thinking about going to 10 mod-54's as my final solution >for this LPAR (roughly 4x the real memory). I wondered what the rest of >you are doing with your paging subsystems in the era of bigass memory sizes. > >Regards, >Tom Conley > >-- > > > >-- >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: Paging subsystems in the era of bigass memory
> -Original Message- > From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On > Behalf Of Tom Marchant > Sent: 12 April, 2017 15:03 > To: IBM-MAIN@LISTSERV.UA.EDU > Subject: Re: Paging subsystems in the era of bigass memory > > On Wed, 12 Apr 2017 06:28:05 +, Vernooij, Kees wrote: > > >From here, there the story still goes on IIRC: if DB2 again needs > >the data, it would be paged in in any normal task. However, DB2 > >is more intelligent: it keeps track of how long it takes to page-in > >the page or read it again from disk, in a I/O that is already running. > >If the latter is faster this is used and the aux slot is really > useless. > > Are you suggesting that before DB2 references a page containing a > buffer, it checks to see if it is paged out? Yes. > And that if it is paged > out, > it doesn't use the record in the buffer, but instead reads it into a > different page? No, it checks its(?) statistics about page-in time and reading it directly from dasd and then decides which will be faster. I heard this years ago, when heavy paging systems might produce slow page-ins and adding a page to an in-progress pre-fetch could be faster. > That makes no sense to me. This does? Kees. > > -- > Tom Marchant For information, services and offers, please visit our web site: http://www.klm.com. This e-mail and any attachment may contain confidential and privileged material intended for the addressee only. If you are not the addressee, you are notified that no part of the e-mail or any attachment may be disclosed, copied or distributed, and that any other action related to this e-mail or attachment is strictly prohibited, and may be unlawful. If you have received this e-mail by error, please notify the sender immediately by return e-mail, and delete this message. Koninklijke Luchtvaart Maatschappij NV (KLM), its subsidiaries and/or its employees shall not be liable for the incorrect or incomplete transmission of this e-mail or any attachments, nor responsible for any delay in receipt. Koninklijke Luchtvaart Maatschappij N.V. (also known as KLM Royal Dutch Airlines) is registered in Amstelveen, The Netherlands, with registered number 33014286 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On Wed, 12 Apr 2017 06:28:05 +, Vernooij, Kees wrote: >From here, there the story still goes on IIRC: if DB2 again needs >the data, it would be paged in in any normal task. However, DB2 >is more intelligent: it keeps track of how long it takes to page-in >the page or read it again from disk, in a I/O that is already running. >If the latter is faster this is used and the aux slot is really useless. Are you suggesting that before DB2 references a page containing a buffer, it checks to see if it is paged out? And that if it is paged out, it doesn't use the record in the buffer, but instead reads it into a different page? That makes no sense to me. -- Tom Marchant -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On Wed, 12 Apr 2017 06:18:29 +, Vernooij, Kees wrote: >The advantage of not freeing the aux slot was that a page could >be paged out without I/O if it had not been changed. Somewhat >the opposite op page reclaim. Freeing it after it has been changed >is of course a 100% useful reclaim of aux slots. Yes, but that would require that someone periodically run through page frames looking for the change bit set, then looking to see if it is backed on AUX, and freeing it if it is. z/OS supports 4 TB of real storage. That's a billion pages to look at. How many CPU cycles is such a reclaim worth? -- Tom Marchant -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
> -Original Message- > From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On > Behalf Of Blaicher, Christopher Y. > Sent: 11 April, 2017 21:25 > To: IBM-MAIN@LISTSERV.UA.EDU > Subject: Re: Paging subsystems in the era of bigass memory > > It has been a while since I worked on DB2, but it is sounding like your > buffer pools are too big. > > Consider this: > DB2 will read a required page into 'new' buffer pool page before it will > invalidate a page it already has in storage. Now we have a physical page > in use. > > The system periodically comes around and looks at the UIC for a page and > if it is high enough, it will page it out. Now we have a page on AUX > storage. > > If DB2 doesn't need the data on that page, or doesn't need to use that > page for a different data page, then that page just hangs out on AUX > storage. > From here, there the story still goes on IIRC: if DB2 again needs the data, it would be paged in in any normal task. However, DB2 is more intelligent: it keeps track of how long it takes to page-in the page or read it again from disk, in a I/O that is already running. If the latter is faster this is used and the aux slot is really useless. Kees. For information, services and offers, please visit our web site: http://www.klm.com. This e-mail and any attachment may contain confidential and privileged material intended for the addressee only. If you are not the addressee, you are notified that no part of the e-mail or any attachment may be disclosed, copied or distributed, and that any other action related to this e-mail or attachment is strictly prohibited, and may be unlawful. If you have received this e-mail by error, please notify the sender immediately by return e-mail, and delete this message. Koninklijke Luchtvaart Maatschappij NV (KLM), its subsidiaries and/or its employees shall not be liable for the incorrect or incomplete transmission of this e-mail or any attachments, nor responsible for any delay in receipt. Koninklijke Luchtvaart Maatschappij N.V. (also known as KLM Royal Dutch Airlines) is registered in Amstelveen, The Netherlands, with registered number 33014286 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
> -Original Message- > From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On > Behalf Of Greg Dyck > Sent: 11 April, 2017 23:39 > To: IBM-MAIN@LISTSERV.UA.EDU > Subject: Re: Paging subsystems in the era of bigass memory > > On 4/11/2017 1:46 PM, Jesse 1 Robinson wrote: > > Part of the problem, I learned some time back at SHARE, is that there > is no mechanism to 'reclaim' page slots that no longer need to remain on > disk. Once storage gets paged out, it sits there like a sandbag until > the owning task is stopped. Contrast that with JES2 spool track reclaim, > which constantly munches through spool like Pacman and frees up unneeded > space. > > Many moons ago code was implemented in ASM and RSM to slowly reclaim ASM > page slots for virtual pages that were changed in real storage. I have > vague memories of this functionality later being disabled for some > technical concern, but can't remember what it was. > > Regards, > Greg > Hi Greg, You still can't resist to jump in on your old hobby of many moons ago? Kees. For information, services and offers, please visit our web site: http://www.klm.com. This e-mail and any attachment may contain confidential and privileged material intended for the addressee only. If you are not the addressee, you are notified that no part of the e-mail or any attachment may be disclosed, copied or distributed, and that any other action related to this e-mail or attachment is strictly prohibited, and may be unlawful. If you have received this e-mail by error, please notify the sender immediately by return e-mail, and delete this message. Koninklijke Luchtvaart Maatschappij NV (KLM), its subsidiaries and/or its employees shall not be liable for the incorrect or incomplete transmission of this e-mail or any attachments, nor responsible for any delay in receipt. Koninklijke Luchtvaart Maatschappij N.V. (also known as KLM Royal Dutch Airlines) is registered in Amstelveen, The Netherlands, with registered number 33014286 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
> > But like any design tradeoff, there are drawbacks as well, and with > real > storage > less expensive and more plentiful, you can have a lot more virtual pages > consuming both an aux slot and a real frame. There has been some > thought given to changing things so that we always free the aux slot > after > we > page-in from it. > > Jim Mulder z/OS Diagnosis, Design, Development, Test IBM Corp. > Poughkeepsie NY > > The advantage of not freeing the aux slot was that a page could be paged out without I/O if it had not been changed. Somewhat the opposite op page reclaim. Freeing it after it has been changed is of course a 100% useful reclaim of aux slots. Kees. For information, services and offers, please visit our web site: http://www.klm.com. This e-mail and any attachment may contain confidential and privileged material intended for the addressee only. If you are not the addressee, you are notified that no part of the e-mail or any attachment may be disclosed, copied or distributed, and that any other action related to this e-mail or attachment is strictly prohibited, and may be unlawful. If you have received this e-mail by error, please notify the sender immediately by return e-mail, and delete this message. Koninklijke Luchtvaart Maatschappij NV (KLM), its subsidiaries and/or its employees shall not be liable for the incorrect or incomplete transmission of this e-mail or any attachments, nor responsible for any delay in receipt. Koninklijke Luchtvaart Maatschappij N.V. (also known as KLM Royal Dutch Airlines) is registered in Amstelveen, The Netherlands, with registered number 33014286 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
l...@garlic.com (Anne & Lynn Wheeler) writes: > The "dup" issue was if aggregate 3880-11 wasn't much larger than > processor memory, then nearly every page in 3880-11 would also in > processor memory. The converse if a page was needed not in processor > memory, then it would unlikely be in 3880-11 cache memory. Moving to > "no-dup" means that a page in processor memory would almost never be > in 3880-11 cache, so there is room for pages (not in processor memory) > that might be needed in processor memory. re: http://www.garlic.com/~lynn/2017d.html#58 Paging subsystems in the era of bigass memory http://www.garlic.com/~lynn/2017d.html#61 Paging subsystems in the era of bigass memory dynamically switching between "dup" and "no-dup" ... "no-dup" was when total processor memory was compareable in size to disk cache for paging (or later, processor memory was compareable in size to space available for disk paging) ... then "no-dup" was sort of like treating/optimizing dynamic disk caches analogous to how the later 3090 extended store was directly treated ... a page was either in processor memory or in extended store ... but wouldn't maintain page in both places. recent posts mentioning 3090 extended store (32M-256N processor storage, 64M-2048M extended storage) http://www.garlic.com/~lynn/2017b.html#69 The ICL 2900 http://www.garlic.com/~lynn/2017d.html#4 GREAT presentation on the history of the mainframe long after all memory could be configured as straight processor memory, LPARs continued to offer option of configuring some as extended store ... apparently because system software had been so structured into supporting processor/extended split ... it took quite some time to adapt to efficiently just using large processor storage. I could also use it for (electronic disk) "1655" ... large number were bought from a vendor for paging use in internal IBM datacenters. It could ran as simulated 2305 (fixed head disk) or as "native" fixed-block device. Larger memory systems ... possibly much larger than total 1655 space, I could dynamically select "no-dup" ... however in the mid-70s I also did "page migration" (which was also included in my resource manager shipped to customers later in the 70s) ... periodically sweaping low-use pages from "fast" paging devices to slower paging devices (but had to be dragged threw memory to move between devices). For 3090 extended store, I wanted to be able to do direct I/O from 3090 extended store (when cleaning pages w/o having to drag through processor memory). (other) recent posts mentioning "1655": http://www.garlic.com/~lynn/2017b.html#68 The ICL 2900 http://www.garlic.com/~lynn/2017c.html#26 Multitasking, together with OS operations posts mentioning page management & replacement algorithms http://www.garlic.com/~lynn/subtopic.html#clock -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
> > If a virtual page is not 'getmain assigned' by VSM it will never be backed > > by a real frame by RSM and an 0C4-11 will occur. When all virtual storage > > on a page is freed VSM will return the allocated AUX slot and real frame, > > if either have been allocated, and reset the 'getmain assigned' indicator. > > > Though maybe not *quite* so dogmatically on a modern processor (a z13, say) > and a modern z/OS... Yes, you can read about that here: http://www-01.ibm.com/support/docview.wss?uid=isg1OA46291 Jim Mulder z/OS Diagnosis, Design, Development, Test IBM Corp. Poughkeepsie NY -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
> What about the other side? Will GETMAIN return indication of success > only if the requested page slots can be committed? There is no consideration of aux slot availability when virtual storage is being allocated. There used to be some capability to reserve aux slots when an address space was created: http://bitsavers.trailing-edge.com/pdf/ibm/370/MVS_XA/GC28-1143-2_MVS-XA_Conversion_Notebook_May84.pdf Using the ASM Backing Slot Function In Release 1.2, the constants that control the number of slots ASM reserves as back up for each new address space or VIO data set are changed to prevent ASM from reserving any. The changes were made because most installations provide adequate paging space and prefer not to use the backing slot function to limit address space and VIO data set creation. If your installation wants ASM to reserve backing slots, you need to change the constants. In Release 1.2, the constants are in the ASMSLOTC and ASMSLOTV fields in the ASJ\1VT. ASM uses the ASMSLOTC value when calculating the number of slots to reserve for address spaces. It uses the ASMSLOTV value when calculating the number for VIO data sets. Earlier releases keep the same values in the nucleus CSECTs ILRSLOTC and ILRSLOTV. Initialization and Tuning describes how ASM uses ASMSLOTC and ASMSLOTV and how to change them. Jim Mulder z/OS Diagnosis, Design, Development, Test IBM Corp. Poughkeepsie NY -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
>Many moons ago code was implemented in ASM and RSM to slowly reclaim ASM >page slots for virtual pages that were changed in real storage. I have >vague memories of this functionality later being disabled for some >technical concern, but can't remember what it was. Slot Scavenger was introduced in OS3/390 2.10. It did its thing during the UIC Update function, UIC Update was removed when Global Steal was implemented (z/OS 1.8, I think), so at that point there was no more Slot Scavenger. Back when real storage was expensive and scarce, then ratio of aux slots to real frames tended to be higher than it is now, and there weren't that many pages that were backed in both real and aux because there wasn't much real. There are some performance advantages to having an unchanged page backed in real and aux. 1. You can steal the frame synchronously, without waiting for page-out I/O. 2. When capturing it during SDUMP, you can give one of the copies to SDUMP with needing to copy it. But like any design tradeoff, there are drawbacks as well, and with real storage less expensive and more plentiful, you can have a lot more virtual pages consuming both an aux slot and a real frame. There has been some thought given to changing things so that we always free the aux slot after we page-in from it. Jim Mulder z/OS Diagnosis, Design, Development, Test IBM Corp. Poughkeepsie NY -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
t...@harminc.net (Tony Harminc) writes: > Not quite sure what you're saying. The old, constrained-memory > technique was usually to issue a variable (Vx) GETMAIN, specifying the > minimum required size as the low bound, and the maximum useful as the > high. Then the system returns the actual amount obtained, or a return > code or abend if it can't deliver even the minimum. In pre-MVS > systems, the REGION= was a hard control on the max; in MVS, REGION= > turned into a per-use limit on Vx GETMAINs, with the hard limit being > the available private area. pre-MVS, ... the application storage management was so bad it required regions typically four times larger than actually used ... typical 1mbyte 370/165 would only run four regions ... and systems were becoming increasingly I/O bound (i.e. CPUs were getting increasingly faster than disks were getting faster, keeping high-end CPUs buzy required lots more concurrent multitasking). justification for moving all 370 to virtual memory was that it would be possible to increase the number of regions on 1mbyte 370/165 by a factor of four times with little or no paging. old post with excerpts with person involved in decision http://www.garlic.com/~lynn/2011d.html#73 Multiple Virtual Memory trivia: I had big argument with the POK people doing the page replacement algorithm. They eventually said it didn't matter anyway ... because there would be almost *NO* paging. It was several years into MVS releases, that somebody "discovers" that side-effect of the implementation, MVS was replacing high-used shared LINKPACK pages before low-used application data pages. past posts mentioning paging algorithm implementations http://www.garlic.com/~lynn/subtopic.html#clock MVS ran into a different kind of problem ... os/360 api paradigm was extensively pointer passing. As a result, an 8mbyte image of the MVS system was included in every application 16mbyte virtual address space (pointer passed in system call and since system code was part of every address space, the old os/360 use of directly using pointer address continued to work). Problem was that subsystem APIs were also pointer based ... and now they were all in their own address space. MVS subsystem calls then invented the common segment area (CSA) that appeared in every address space and was used for allocating space for subsystem API parameters (reducing applications space to 7mbytes, out of 16mbytes). However, CSA size requirements were somewhat proportional to the number of subsystems and the number of concurrent applications and CSA frequently became multiple segments and morphed into common system area. By 3033, CSA space requirements was frequently 5-6mbytes for many customers (leaving only 2mbytes for applications) and threatening to become 8mbytes (leaving zero bytes for applications) Eventually part of 370-xa access registers were retrofitted to 370 as dual-address space mode subsystems could be enhanced to directly access application space w/o requiring CSA (person responsible for dual-access retrofit leaves not long after for HP to work on their risc processors). other triva: early 80s I was saying that disk performance not tracking was so bad that since the 60s, relative disk system throughput had declined by a factor of ten times (i.e. disks got 3-5 times faster, processors got 50 times faster). disk division took exception and assigned division performance organization to refute my claims. after a few weeks, they came back and basically said that I had understated the "problem". They respin the analysis into SHARE presentation (B874 at SHARE 63) recommending disk configurations to improve throughput. old post with part of the early 80 comparison http://www.garlic.com/~lynn/93.html#31 old posts with pieces of B874 http://www.garlic.com/~lynn/2001l.html#56 http://www.garlic.com/~lynn/2006f.html#3 for 3081 & 3880-11 paging caches ... processor memory were becoming compareable in size ... or even larger than paging area ... so I did a variation that dynamically switched between what I called "dup" (duplicate) and "no-dup". "dup", when page was read into processor memory, the original was left allocated (duplicate in memory and on disk), if that page was later replaced and had not been changed, then it could just be invalidated (and didn't have to be written out) since an exact copy was already on disk. For large processor memory, it could dynamically switch to "no-dup" and read into processor would always deallocate the copy on disk (and would use 3880-11 no-cache read, if a copy was in cache, it would be read and removed from cache, if not in cache, it would read from disk bypassing cache). The "dup" issue was if aggregate 3880-11 wasn't much larger than processor memory, then nearly every page in 3880-11 would also in processor memory. The converse if a page was needed not in processor memory, then it would unlikely be in 3880-11 cache memory. Moving to "no-dup" means that a page in processor me
Re: Paging subsystems in the era of bigass memory
On Tue, 11 Apr 2017 16:45:10 -0500, Greg Dyck wrote: >On 4/11/2017 3:26 PM, Paul Gilmartin wrote: >> My understanding, ancient, probably outdated, and certainly naive is that >> there is little communication between GETMAIN/FREEMAIN and the paging >> subsystem. If a program touches a page that was never GETMAINed no error >> occurs; simply a page slot is allocated up to the limit of the REGION >> parameter. Conversely, on FREEMAIN the page slots are not released. > >*Totally* wrong. z/OS (or any predecessors or relatives like VS1) has >never worked like you describe. > >If a virtual page is not 'getmain assigned' by VSM it will never be >backed by a real frame by RSM and an 0C4-11 will occur. When all >virtual storage on a page is freed VSM will return the allocated AUX >slot and real frame, if either have been allocated, and reset the >'getmain assigned' indicator. > What about the other side? Will GETMAIN return indication of success only if the requested page slots can be committed? A Google search for "lazy malloc" shows the preponderance of complaints concern Linux. My information came from a UNIX/C programmer before the ascendancy of Linux. And from an MVS wizard who insisted that for GETMAIN to fail if requested page slots could not be committed would cause enormous breakage of existing art. -- gil -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 11 April 2017 at 17:45, Greg Dyck wrote: > If a virtual page is not 'getmain assigned' by VSM it will never be backed > by a real frame by RSM and an 0C4-11 will occur. When all virtual storage > on a page is freed VSM will return the allocated AUX slot and real frame, > if either have been allocated, and reset the 'getmain assigned' indicator. Though maybe not *quite* so dogmatically on a modern processor (a z13, say) and a modern z/OS... Tony H. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 4/11/2017 3:26 PM, Paul Gilmartin wrote: My understanding, ancient, probably outdated, and certainly naive is that there is little communication between GETMAIN/FREEMAIN and the paging subsystem. If a program touches a page that was never GETMAINed no error occurs; simply a page slot is allocated up to the limit of the REGION parameter. Conversely, on FREEMAIN the page slots are not released. *Totally* wrong. z/OS (or any predecessors or relatives like VS1) has never worked like you describe. If a virtual page is not 'getmain assigned' by VSM it will never be backed by a real frame by RSM and an 0C4-11 will occur. When all virtual storage on a page is freed VSM will return the allocated AUX slot and real frame, if either have been allocated, and reset the 'getmain assigned' indicator. Regards, Greg -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 4/11/2017 1:46 PM, Jesse 1 Robinson wrote: Part of the problem, I learned some time back at SHARE, is that there is no mechanism to 'reclaim' page slots that no longer need to remain on disk. Once storage gets paged out, it sits there like a sandbag until the owning task is stopped. Contrast that with JES2 spool track reclaim, which constantly munches through spool like Pacman and frees up unneeded space. Many moons ago code was implemented in ASM and RSM to slowly reclaim ASM page slots for virtual pages that were changed in real storage. I have vague memories of this functionality later being disabled for some technical concern, but can't remember what it was. Regards, Greg -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 4/11/2017 2:24 PM, Blaicher, Christopher Y. wrote: It has been a while since I worked on DB2, but it is sounding like your buffer pools are too big. With large memory systems everyone should have all of the production (and, ideally, test systems too) buffer pools define with PGFIX(YES) specified. Even if they are over defined in size. This can *measurably* cut DB2 CPU overhead by eliminating the need to fix and unfix pages for I/O and buffer pool pages will not unexpectedly creep out to AUX if they are not changed for a long time. Regards, Greg -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 11 April 2017 at 16:26, Paul Gilmartin < 000433f07816-dmarc-requ...@listserv.ua.edu> wrote: > If a program touches a page that was never GETMAINed no error > occurs; simply a page slot is allocated up to the limit of the REGION > parameter. Conversely, on FREEMAIN the page slots are not released. > I really don't think so... It's easy to experiment. > > Programmers with a strong UNIX/C background loathe this. > Maybe because they believe in all kinds of mainframe myths. The machines are run by nerdy looking men in white lab coats, with ties mandatory, I've heard. And everything requires punch cards to get any work done. They are accustomed to do needed malloc()s up front and handle errors at > that > point; their code is not designed to handle out-of-storage conditions > that occur unexpectedly, later when a page is touched. > Sounds like, uh z/OS. > > From the other point of view, traditional OS programmers are accustomed > to code enormous fixed GETMAINs in case they need the storage years in > the future -- just adjust the REGION and the program works again. > Not quite sure what you're saying. The old, constrained-memory technique was usually to issue a variable (Vx) GETMAIN, specifying the minimum required size as the low bound, and the maximum useful as the high. Then the system returns the actual amount obtained, or a return code or abend if it can't deliver even the minimum. In pre-MVS systems, the REGION= was a hard control on the max; in MVS, REGION= turned into a per-use limit on Vx GETMAINs, with the hard limit being the available private area. In more recent systems there are exits that can limit most of these values, and of course there is split BTL and ATL storage, not to mention ATB. > > Culture clash. > Perhaps, but it's unwise to ascribe to the "other" culture incorrect facts. Tony H. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 2017-04-11, at 12:47, Jesse 1 Robinson wrote: > > Part of the problem, I learned some time back at SHARE, is that there is no > mechanism to 'reclaim' page slots that no longer need to remain on disk. Once > storage gets paged out, it sits there like a sandbag until the owning task is > stopped. ... > My understanding, ancient, probably outdated, and certainly naive is that there is little communication between GETMAIN/FREEMAIN and the paging subsystem. If a program touches a page that was never GETMAINed no error occurs; simply a page slot is allocated up to the limit of the REGION parameter. Conversely, on FREEMAIN the page slots are not released. Programmers with a strong UNIX/C background loathe this. They are accustomed to do needed malloc()s up front and handle errors at that point; their code is not designed to handle out-of-storage conditions that occur unexpectedly, later when a page is touched. From the other point of view, traditional OS programmers are accustomed to code enormous fixed GETMAINs in case they need the storage years in the future -- just adjust the REGION and the program works again. Culture clash. Repairing this to make paging substem aware of FREEMAIN and release the slots would be an enormous and risky design change. -- gil -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
Chris I think that's entirely possible, I am no DB2 person, but I have installed and configured tools for DB2 change data capture, DB2 to UDB, DB2 to SQL, and I think the DBA's at least my DBA's define the buffer pools larger for that purpose? I also have large frame area defined of 7740M Carmen From: "Christopher Y. Blaicher" To: IBM-MAIN@LISTSERV.UA.EDU Sent: Tuesday, April 11, 2017 2:24:54 PM Subject: Re: Paging subsystems in the era of bigass memory It has been a while since I worked on DB2, but it is sounding like your buffer pools are too big. Consider this: DB2 will read a required page into 'new' buffer pool page before it will invalidate a page it already has in storage. Now we have a physical page in use. The system periodically comes around and looks at the UIC for a page and if it is high enough, it will page it out. Now we have a page on AUX storage. If DB2 doesn't need the data on that page, or doesn't need to use that page for a different data page, then that page just hangs out on AUX storage. I don't know what happens if you close a tablespace. DB2 probably just frees its logical use, but doesn't FREEMAIN the storage. Bottom line, you basically need enough AUX storage to hold all your buffer pools. If you have a lot of pages just hanging out in AUX and you don't have any demand paging, maybe you have buffer pools a little larger than you need. Of course, the DB2 guys want a minimum of reads because a read is much slower than a page-in, but that is something for individual shops to work out. Joel Goldstein can probably wax poetic on this topic much more than I can. Chris Blaicher Technical Architect Mainframe Development Syncsort Incorporated 2 Blue Hill Plaza #1563, Pearl River, NY 10965 P: 201-930-8234 | M: 512-627-3803 E: cblaic...@syncsort.com www.syncsort.com CONNECTING BIG IRON TO BIG DATA -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Jesse 1 Robinson Sent: Tuesday, April 11, 2017 2:47 PM To: IBM-MAIN@LISTSERV.UA.EDU Subject: Re: Paging subsystems in the era of bigass memory The problem we face is 'paging creep'. Right after IPL, systems show 0% ASM usage for some period of time. Then usage starts to creep up until we get warnings, then eventually hit the no-more-SVC-dumps condition. Adding memory to an LPAR slows the creep but cannot seem to stop it altogether. The problem is most pronounced on systems with large DB2 apps. Part of the problem, I learned some time back at SHARE, is that there is no mechanism to 'reclaim' page slots that no longer need to remain on disk. Once storage gets paged out, it sits there like a sandbag until the owning task is stopped. Contrast that with JES2 spool track reclaim, which constantly munches through spool like Pacman and frees up unneeded space. . . 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 robin...@sce.com -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Tom Conley Sent: Tuesday, April 11, 2017 11:13 AM To: IBM-MAIN@LISTSERV.UA.EDU Subject: (External):Re: Paging subsystems in the era of bigass memory On 4/11/2017 1:16 PM, van der Grijn, Bart , B wrote: > Largest LPARs we have are about 200GB with 6 MOD27 per LPAR. They all run DB2 > for distributed workloads plus some application specific subsystems. > The two busiest of those LPARs each run one DB2 member of the same DB2 data > sharing group with a frame occupancy of about 39M. > Next to no paging. > > Bart > Bart, This is what has me puzzled. My two biggest users of AUX, according to TMONMVS, are our two DB2 production regions. They're like 90% of what's in the page datasets. I have the DB2 sysprog looking at DB2's virtual storage knobs to see if we have something misconfigured. Thanks, Tom Conley -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN ATTENTION: - The information contained in this message (including any files transmitted with this message) may contain proprietary, trade secret or other confidential and/or legally privileged information. Any pricing information contained in this message or in any files transmitted with this message is always confidential and cannot be shared with any third parties without prior written approval from Syncsort. This message is intended to be read only by the individual or entity to whom it is addressed or by their designee. If the reader of this message is not the in
Re: Paging subsystems in the era of bigass memory
It has been a while since I worked on DB2, but it is sounding like your buffer pools are too big. Consider this: DB2 will read a required page into 'new' buffer pool page before it will invalidate a page it already has in storage. Now we have a physical page in use. The system periodically comes around and looks at the UIC for a page and if it is high enough, it will page it out. Now we have a page on AUX storage. If DB2 doesn't need the data on that page, or doesn't need to use that page for a different data page, then that page just hangs out on AUX storage. I don't know what happens if you close a tablespace. DB2 probably just frees its logical use, but doesn't FREEMAIN the storage. Bottom line, you basically need enough AUX storage to hold all your buffer pools. If you have a lot of pages just hanging out in AUX and you don't have any demand paging, maybe you have buffer pools a little larger than you need. Of course, the DB2 guys want a minimum of reads because a read is much slower than a page-in, but that is something for individual shops to work out. Joel Goldstein can probably wax poetic on this topic much more than I can. Chris Blaicher Technical Architect Mainframe Development Syncsort Incorporated 2 Blue Hill Plaza #1563, Pearl River, NY 10965 P: 201-930-8234 | M: 512-627-3803 E: cblaic...@syncsort.com www.syncsort.com CONNECTING BIG IRON TO BIG DATA -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Jesse 1 Robinson Sent: Tuesday, April 11, 2017 2:47 PM To: IBM-MAIN@LISTSERV.UA.EDU Subject: Re: Paging subsystems in the era of bigass memory The problem we face is 'paging creep'. Right after IPL, systems show 0% ASM usage for some period of time. Then usage starts to creep up until we get warnings, then eventually hit the no-more-SVC-dumps condition. Adding memory to an LPAR slows the creep but cannot seem to stop it altogether. The problem is most pronounced on systems with large DB2 apps. Part of the problem, I learned some time back at SHARE, is that there is no mechanism to 'reclaim' page slots that no longer need to remain on disk. Once storage gets paged out, it sits there like a sandbag until the owning task is stopped. Contrast that with JES2 spool track reclaim, which constantly munches through spool like Pacman and frees up unneeded space. . . 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 robin...@sce.com -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Tom Conley Sent: Tuesday, April 11, 2017 11:13 AM To: IBM-MAIN@LISTSERV.UA.EDU Subject: (External):Re: Paging subsystems in the era of bigass memory On 4/11/2017 1:16 PM, van der Grijn, Bart , B wrote: > Largest LPARs we have are about 200GB with 6 MOD27 per LPAR. They all run DB2 > for distributed workloads plus some application specific subsystems. > The two busiest of those LPARs each run one DB2 member of the same DB2 data > sharing group with a frame occupancy of about 39M. > Next to no paging. > > Bart > Bart, This is what has me puzzled. My two biggest users of AUX, according to TMONMVS, are our two DB2 production regions. They're like 90% of what's in the page datasets. I have the DB2 sysprog looking at DB2's virtual storage knobs to see if we have something misconfigured. Thanks, Tom Conley -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN ATTENTION: - The information contained in this message (including any files transmitted with this message) may contain proprietary, trade secret or other confidential and/or legally privileged information. Any pricing information contained in this message or in any files transmitted with this message is always confidential and cannot be shared with any third parties without prior written approval from Syncsort. This message is intended to be read only by the individual or entity to whom it is addressed or by their designee. If the reader of this message is not the intended recipient, you are on notice that any use, disclosure, copying or distribution of this message, in any form, is strictly prohibited. If you have received this message in error, please immediately notify the sender and/or Syncsort and destroy all copies of this message in your possession, custody or control. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 4/11/2017 8:54 AM, Doug Henry wrote: Hi Tom, We have a pair of SCM cards (so in case one fails). Each card has 1.4 Terabytes of memory. For an lpar that has 524GB online and the SCM online is 384G. As for the cost I don't really know. ISTR a pair of flash cards is somewhere around $150K. -- Edward E Jaffe Phoenix Software International, Inc 831 Parkview Drive North El Segundo, CA 90245 http://www.phoenixsoftware.com/ -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
Right there with you - last IPL of the prod system was Feb 26th, all LOCAL page datasets are sitting @ 20%, I was told this was NORMAL for DB2 - Original Message - From: "Tom Conley" To: IBM-MAIN@LISTSERV.UA.EDU Sent: Tuesday, April 11, 2017 1:52:06 PM Subject: Re: Paging subsystems in the era of bigass memory On 4/11/2017 2:46 PM, Jesse 1 Robinson wrote: > The problem we face is 'paging creep'. Right after IPL, systems show 0% ASM > usage for some period of time. Then usage starts to creep up until we get > warnings, then eventually hit the no-more-SVC-dumps condition. Adding memory > to an LPAR slows the creep but cannot seem to stop it altogether. The problem > is most pronounced on systems with large DB2 apps. > > Part of the problem, I learned some time back at SHARE, is that there is no > mechanism to 'reclaim' page slots that no longer need to remain on disk. Once > storage gets paged out, it sits there like a sandbag until the owning task is > stopped. Contrast that with JES2 spool track reclaim, which constantly > munches through spool like Pacman and frees up unneeded space. > > . > . > 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 > robin...@sce.com > Preach it Brother Skip!! After IPL, I'm at 0%. A week in, 15%, a month in, 44%. If this is indeed a "feature", then we need to create a requirement to fix this. Tom -- 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: EXTERNAL: Re: Paging subsystems in the era of bigass memory
Certainly sounds like requirement time. We see the same issue particularly on one of our Dev systems with 14 DB2 members on it. Jerry Whitteridge Manager Mainframe Systems & Storage Albertsons - Safeway Inc. 623 869 5523 Corporate Tieline - 85523 If you feel in control you just aren't going fast enough. -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Tom Conley Sent: Tuesday, April 11, 2017 11:52 AM To: IBM-MAIN@LISTSERV.UA.EDU Subject: EXTERNAL: Re: Paging subsystems in the era of bigass memory On 4/11/2017 2:46 PM, Jesse 1 Robinson wrote: > The problem we face is 'paging creep'. Right after IPL, systems show 0% ASM > usage for some period of time. Then usage starts to creep up until we get > warnings, then eventually hit the no-more-SVC-dumps condition. Adding memory > to an LPAR slows the creep but cannot seem to stop it altogether. The problem > is most pronounced on systems with large DB2 apps. > > Part of the problem, I learned some time back at SHARE, is that there is no > mechanism to 'reclaim' page slots that no longer need to remain on disk. Once > storage gets paged out, it sits there like a sandbag until the owning task is > stopped. Contrast that with JES2 spool track reclaim, which constantly > munches through spool like Pacman and frees up unneeded space. > > . > . > 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 > robin...@sce.com > Preach it Brother Skip!! After IPL, I'm at 0%. A week in, 15%, a month in, 44%. If this is indeed a "feature", then we need to create a requirement to fix this. Tom -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN Warning: All e-mail sent to this address will be received by the corporate e-mail system, and is subject to archival and review by someone other than the recipient. This e-mail may contain proprietary information and is intended only for the use of the intended recipient(s). If the reader of this message is not the intended recipient(s), you are notified that you have received this message in error and that any review, dissemination, distribution or copying of this message is strictly prohibited. If you have received this message in error, please notify the sender immediately. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 4/11/2017 2:46 PM, Jesse 1 Robinson wrote: The problem we face is 'paging creep'. Right after IPL, systems show 0% ASM usage for some period of time. Then usage starts to creep up until we get warnings, then eventually hit the no-more-SVC-dumps condition. Adding memory to an LPAR slows the creep but cannot seem to stop it altogether. The problem is most pronounced on systems with large DB2 apps. Part of the problem, I learned some time back at SHARE, is that there is no mechanism to 'reclaim' page slots that no longer need to remain on disk. Once storage gets paged out, it sits there like a sandbag until the owning task is stopped. Contrast that with JES2 spool track reclaim, which constantly munches through spool like Pacman and frees up unneeded space. . . 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 robin...@sce.com Preach it Brother Skip!! After IPL, I'm at 0%. A week in, 15%, a month in, 44%. If this is indeed a "feature", then we need to create a requirement to fix this. Tom -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
The problem we face is 'paging creep'. Right after IPL, systems show 0% ASM usage for some period of time. Then usage starts to creep up until we get warnings, then eventually hit the no-more-SVC-dumps condition. Adding memory to an LPAR slows the creep but cannot seem to stop it altogether. The problem is most pronounced on systems with large DB2 apps. Part of the problem, I learned some time back at SHARE, is that there is no mechanism to 'reclaim' page slots that no longer need to remain on disk. Once storage gets paged out, it sits there like a sandbag until the owning task is stopped. Contrast that with JES2 spool track reclaim, which constantly munches through spool like Pacman and frees up unneeded space. . . 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 robin...@sce.com -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Tom Conley Sent: Tuesday, April 11, 2017 11:13 AM To: IBM-MAIN@LISTSERV.UA.EDU Subject: (External):Re: Paging subsystems in the era of bigass memory On 4/11/2017 1:16 PM, van der Grijn, Bart , B wrote: > Largest LPARs we have are about 200GB with 6 MOD27 per LPAR. They all run DB2 > for distributed workloads plus some application specific subsystems. > The two busiest of those LPARs each run one DB2 member of the same DB2 data > sharing group with a frame occupancy of about 39M. > Next to no paging. > > Bart > Bart, This is what has me puzzled. My two biggest users of AUX, according to TMONMVS, are our two DB2 production regions. They're like 90% of what's in the page datasets. I have the DB2 sysprog looking at DB2's virtual storage knobs to see if we have something misconfigured. Thanks, Tom Conley -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 4/11/2017 1:16 PM, van der Grijn, Bart , B wrote: Largest LPARs we have are about 200GB with 6 MOD27 per LPAR. They all run DB2 for distributed workloads plus some application specific subsystems. The two busiest of those LPARs each run one DB2 member of the same DB2 data sharing group with a frame occupancy of about 39M. Next to no paging. Bart Bart, This is what has me puzzled. My two biggest users of AUX, according to TMONMVS, are our two DB2 production regions. They're like 90% of what's in the page datasets. I have the DB2 sysprog looking at DB2's virtual storage knobs to see if we have something misconfigured. Thanks, Tom Conley -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
Largest LPARs we have are about 200GB with 6 MOD27 per LPAR. They all run DB2 for distributed workloads plus some application specific subsystems. The two busiest of those LPARs each run one DB2 member of the same DB2 data sharing group with a frame occupancy of about 39M. Next to no paging. Bart -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Pinnacle Sent: Tuesday, April 11, 2017 10:47 AM To: IBM-MAIN@LISTSERV.UA.EDU Subject: Paging subsystems in the era of bigass memory Gone are the halcyon days when we could run an LPAR with three mod-3's as the local paging subsystem. With today's large memory sizes, I'm faced with having to completely rethink my paging subsystems. I've currently got a 133GB LPAR with 18 mod-9 locals at 44%. I'm going to add 22 more mod-9's, which will get me just under the 30% threshold. That's 40 page datasets, which is about 30 more than the most I've ever managed. I'm thinking about going to 10 mod-54's as my final solution for this LPAR (roughly 4x the real memory). I wondered what the rest of you are doing with your paging subsystems in the era of bigass memory sizes. Regards, Tom Conley -- -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
Don't forget about PAGTOTL in SYS1.PARMLIB(IEASYS00). Default is 40. Extra slots reserve ESQA, so don't be overly generous. HTH, I had the total number of pages allocated and available in the current paging subsystem, so I calculated the available pages with 22 more mod-9's and then divided the allocated pages into the new total pages available. I plan to do PAGEADDs for the new page datasets, and then PAGEDEL and PAGEADD the existing datasets to redistribute the pages. ::DISCLAIMER:: The contents of this e-mail and any attachment(s) are confidential and intended for the named recipient(s) only. E-mail transmission is not guaranteed to be secure or error-free as information could be intercepted, corrupted, lost, destroyed, arrive late or incomplete, or may contain viruses in transmission. The e mail and its contents (with or without referred errors) shall therefore not attach any liability on the originator or HCL or its affiliates. Views or opinions, if any, presented in this email are solely those of the author and may not necessarily reflect the views or opinions of HCL or its affiliates. Any form of reproduction, dissemination, copying, disclosure, modification, distribution and / or publication of this message without the prior written consent of authorized representative of HCL is strictly prohibited. If you have received this email in error please delete it and notify the sender immediately. Before opening any email and/or attachments, please check them for viruses and other defects. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On Tue, 11 Apr 2017 12:36:10 -0400, Tom Conley wrote: >I had the total number of pages allocated and available in the current >paging subsystem, so I calculated the available pages with 22 more >mod-9's and then divided the allocated pages into the new total pages >available. I plan to do PAGEADDs for the new page datasets, and then >PAGEDEL and PAGEADD the existing datasets to redistribute the pages. > >Regards, >Tom Conley > Better to use PAGEDEL REPLACE for the ones that will be replaced. Then maybe wait for an IPL if you can to eliminate the others after changing IEASYSxx.See the warning in the MVS commands manual. Yes, I ran into that problem many years ago when going from mod-3 to mod-9 local page data sets. Regards, Mark -- Mark Zelden - Zelden Consulting Services - z/OS, OS/390 and MVS ITIL v3 Foundation Certified mailto:m...@mzelden.com Mark's MVS Utilities: http://www.mzelden.com/mvsutil.html Systems Programming expert at http://search390.techtarget.com/ateExperts/ -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 4/11/2017 12:03 PM, Lizette Koehler wrote: Tom How did you determine you needed 22 more MOD9s. Is there a formula you used? Lizette Hi Lizette, I had the total number of pages allocated and available in the current paging subsystem, so I calculated the available pages with 22 more mod-9's and then divided the allocated pages into the new total pages available. I plan to do PAGEADDs for the new page datasets, and then PAGEDEL and PAGEADD the existing datasets to redistribute the pages. Regards, Tom Conley -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On Tue, 11 Apr 2017 10:46:40 -0400, Pinnacle wrote: >Gone are the halcyon days when we could run an LPAR with three mod-3's >as the local paging subsystem. With today's large memory sizes, I'm >faced with having to completely rethink my paging subsystems. I've >currently got a 133GB LPAR with 18 mod-9 locals at 44%. I'm going to >add 22 more mod-9's, which will get me just under the 30% threshold. >That's 40 page datasets, which is about 30 more than the most I've ever >managed. I'm thinking about going to 10 mod-54's as my final solution >for this LPAR (roughly 4x the real memory). I wondered what the rest of >you are doing with your paging subsystems in the era of bigass memory sizes. > >Regards, >Tom Conley > The 3 largest LPARs in terms of memory in one of the sysplexes I support are: 286G, running dev WebSphere and 2 prod WAS LPARs each with 230G. I know there is not enough locals to cover all these 64-bit workloads actually using what their potential is - but then again, I don't have enough DASD of that in the entire DASD farm. We try not to page on those prod LPARs so there is enough real storage to support that and the local page DS usage is near 0%. There are 9 3390-27 locals on those 2 prod LPARs. OTOH, the dev LPAR with 286G has a larger virtual storage requirement due to so many dev WAS regions. There are 16 mod 27s for locals and right now the usage is 6%. Other LPARs in the sysplex running CICS, DB2, MQ and most of of the batch have an average of about 80G of real storage and 7-10 locals, all 3390 mod 27s. There is no way you could possible plan for worse case scenarios with 64-bit. I think you have to look at what your actual workloads are and what happens during "bad" events (like multiple SVC dumps) to have extra storage available for that so paging doesn't go through the roof. (I wish we had flash in our z13s for SVC dumps, but we don't). We do keep page ds usage at 30% or lower per the old ROT and I still see that ROT sited. But honestly, I don't get it... with mod-27s and mod-54s for locals, how could ASM not find enough contiguous slots to be efficient even if the page DS is 50% utilized. Best Regards, Mark -- Mark Zelden - Zelden Consulting Services - z/OS, OS/390 and MVS ITIL v3 Foundation Certified mailto:m...@mzelden.com Mark's MVS Utilities: http://www.mzelden.com/mvsutil.html Systems Programming expert at http://search390.techtarget.com/ateExperts/ -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
Tom How did you determine you needed 22 more MOD9s. Is there a formula you used? Lizette > -Original Message- > From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On > Behalf Of Pinnacle > Sent: Tuesday, April 11, 2017 7:47 AM > To: IBM-MAIN@LISTSERV.UA.EDU > Subject: Paging subsystems in the era of bigass memory > > Gone are the halcyon days when we could run an LPAR with three mod-3's as the > local paging subsystem. With today's large memory sizes, I'm faced with > having to completely rethink my paging subsystems. I've currently got a 133GB > LPAR with 18 mod-9 locals at 44%. I'm going to add 22 more mod-9's, which > will get me just under the 30% threshold. > That's 40 page datasets, which is about 30 more than the most I've ever > managed. I'm thinking about going to 10 mod-54's as my final solution for this > LPAR (roughly 4x the real memory). I wondered what the rest of you are doing > with your paging subsystems in the era of bigass memory sizes. > > Regards, > Tom Conley > > -- > > > > -- > 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: Paging subsystems in the era of bigass memory
On Tue, 11 Apr 2017 11:31:32 -0400, Tom Conley wrote: >On 4/11/2017 11:20 AM, Doug Henry wrote: >> On Tue, 11 Apr 2017 10:46:40 -0400, Pinnacle >> wrote: >> >> I wondered what the rest of >>> you are doing with your paging subsystems in the era of bigass memory sizes. >>> >>> Regards, >>> Tom Conley >> Hi Tom, >> We use Storage Class Memory to solve the problem. >> >> Doug Henry >> USBANK >> > >Hey Doug, > >Oh, a wiseguy (nyuk nyuk). How much real memory and how much SCM? And >is the price point for SCM competitive with a similar amount of DASD? >It might be if you go with the old 2-3X real (e.g. 4-6TB of DASD just to >back 2TB real). > >Regards, >Tom Conley > Hi Tom, We have a pair of SCM cards (so in case one fails). Each card has 1.4 Terabytes of memory. For an lpar that has 524GB online and the SCM online is 384G. As for the cost I don't really know. Doug -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On 4/11/2017 11:20 AM, Doug Henry wrote: On Tue, 11 Apr 2017 10:46:40 -0400, Pinnacle wrote: I wondered what the rest of you are doing with your paging subsystems in the era of bigass memory sizes. Regards, Tom Conley Hi Tom, We use Storage Class Memory to solve the problem. Doug Henry USBANK Hey Doug, Oh, a wiseguy (nyuk nyuk). How much real memory and how much SCM? And is the price point for SCM competitive with a similar amount of DASD? It might be if you go with the old 2-3X real (e.g. 4-6TB of DASD just to back 2TB real). Regards, Tom Conley -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Paging subsystems in the era of bigass memory
On Tue, 11 Apr 2017 10:46:40 -0400, Pinnacle wrote: I wondered what the rest of >you are doing with your paging subsystems in the era of bigass memory sizes. > >Regards, >Tom Conley Hi Tom, We use Storage Class Memory to solve the problem. Doug Henry USBANK -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN