Re: Where to find statistics for fetched programs?
On Wed, 7 Oct 2009 13:56:48 -0500, Jan Vanbrabant wrote: But I couldn't dig up what I was looking for: will one of the reports display the CPU usage within each module? Jan, None of the reports you listed will show CPU consumption by program, because that information is not collected. Reported usage for a program will be the same whether it consumed a CPU minute or a CPU hour. Now, the reported usage will be different if that CPU minute (or hour) is spread over one invocation or one thousand invocations, but only in terms of use count. If those thousand invocations are from the one job, then that only counts as one job using the program. This is the sort of situation where the choice of report may become significant. [Note that the above is based Dorana/TAD4z whereas I will not attest to how usage reported by TLCMz is affected by these illustrative comparisons. It is possible that it is slightly different in terms of usage counts.] You're right that Dorana/TLCMz/TAD4z will report the use of programs even if they are not named on EXEC JCL statements. LLA exits can also do this, but only for programs fetched from LLA-managed libraries. I was thinking of a scheme of monitoring CPU time at the RB level back in the Dorana days if there was any interest. There wasn't. The idea was that the CPU time of an RB would be accumulated and logged against the name of the program that the RB pointed to. Note this this means that CPU time spent in programs that were simply branched to would be logged against the programs that branched to them. The scheme would have been strictly for TCB mode only, and so these days may be considered incomplete, and therefore misleading to some extent. Still, it never got beyond the idea stage, so I guess the technical issues are not a problem. Cheers, Greg -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
Hi Greg, Re. you saying ... you could check out the recently GA-ed ‘’Tivoli Asset Discovery for z/OS’’ which happens to employ some of the Dorana technology I helped write way back last millenium. This product is also the logical successor to TLCMz, if I'm not mistaken. Yes , you're right: ‘Tivoli Asset Discovery for z/OS’ (TAD4z) replaces ‘Tivoli License Compliance Manager for z/OS’ (TLCMz). I've been reading TAD4z 's InfoCenter http://publib.boulder.ibm.com/infocenter/tivihelp/v29r1/index.jsp SC27-2701-00 IBM Tivoli Asset Discovery for z/OS V7.2 Administration Guide and Reference (August 2009) http://publib.boulder.ibm.com/infocenter/tivihelp/v29r1/topic/com.ibm.tad4z.do c/c2727010.pdf But I couldn't dig up what I was looking for: will one of the reports display the CPU usage within each module? Or will it just help to identify those modules that are invoked by non-EXEC statement means? Like tour LLA exit did (still does) back last millenium? Where can I find the details (contents) of the TAD4z reports? I only find their names in Table 15. Tivoli Asset Discovery for z/OS ad hoc reports Member Description HSISIPIDINSTALLED-PRODUCT-IDENTIFICATION-DETAIL HSISIPV INSTALLED-PRODUCT-SUMMARY HSISLLIDLOAD-LIBRARY-IDENTIFICATION-DETAIL HSISLLS LOAD-LIBRARY-SUMMARY HSISLMD LOAD-MODULE-DETAIL HSISMUSLMODULE-USE-SUMMARY-BY-LIBRARY-AND-PRODUCT HSISMUSPMODULE-USE-SUMMARY-BY-PRODUCT-AND-LIBRARY HSISPDLPPRODUCT-USE-DETAIL-BY-LIBRARY-AND-PRODUCT HSISPDPLPRODUCT-USE-DETAIL-BY-PRODUCT-AND-LIBRARY HSISPJD PRODUCT-USE-DETAIL-BY-JOBNAME-PRODUCT--LIBRARY HSISPJP PRODUCT-USE-SUMMARY-BY-JOBNAME-AND-PRODUCT HSISPJPLPRODUCT-USE-SUMMARY-BY-JOBNAME-PRODUCT-- LIBRARY HSISPSLPPRODUCT-USE-SUMMARY-BY-LIBRARY-AND-PRODUCT HSISPSP PROD-USE-SUMMARY-BY-PRODUCT HSISPSPLPRODUCT-USE-SUMMARY-BY-PRODUCT-AND-LIBRARY HSISPUD PRODUCT-USE-DETAIL-BY-USERID-PRODUCT-AND-LIBRARY HSISPUP PRODUCT-USE-SUMMARY-BY-USERID-AND-PRODUCT HSISPUPLPRODUCT-USE-SUMMARY-BY-USERID-PRODUCT-AND- LIBRARY HSISPV PRODUCTS-BY-VENDOR HSISVS VERSION-SUMMARY HSISLINVList all inventories in DB2 Repository Though involved with the Dorana technology, you seem not be involved anymore in TAD4z, so you might not know the answer to this. But you might possibly route me to appropriate contact points or additional points of information. Cheers, Jan -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
snip That is very true. One side issue is that every once in a while there is a program that *NOBODY* knows about and you have done your home work researching where it might be called from. You cannot find any obvious answer so after notifying everyone you delete the module and all of a sudden a lot of programs abend with an S806 etc... You quickly put it back and everything is back to normal, meanwhile you have egg on your face. That has been the issue since day 1 of of os/360 and it is still valid today. Unless you have a front end to fetch and keep your own numbers you are playing craps with possibility. Personally I would love to know via SMF say a record is cut with totals every so often (user specifiable) that would help a lot. It would be nice to have other information as well (library found in or link list library number). It seems a reasonable thing to do maybe some one has written a SHARE requirement or should submit one. -unsnip-- I know of a LLA exit written by Greg Price that can help with module counts. Perhaps he could be persuaded to release that exit, along with a batch program to fetch/display the counts and reset the counters. I have used this exit and one reporting mechanism in the past and it seems to work very well. It won't display the CPU usage within each module, but it will at least help to idntify those modules that are invoked by non-EXEC statement means. I cannot share it because of various agreements with Greg, but perhaps he can be persuaded to make something of this nature for a very nominal fee, or perhaps free. (I must point out that every scrap of code that I've ever used from Greg has been letter-perfect and completely defect free.) Rick -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
Does it also tell you what program invoked the module? Joel Wolpert Performance and Capacity Planning consultant WEBSITE: www.perfconsultant.com - Original Message - From: Rick Fochtman rfocht...@ync.net Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@bama.ua.edu Sent: Thursday, August 20, 2009 12:49 PM Subject: Re: Where to find statistics for fetched programs? snip That is very true. One side issue is that every once in a while there is a program that *NOBODY* knows about and you have done your home work researching where it might be called from. You cannot find any obvious answer so after notifying everyone you delete the module and all of a sudden a lot of programs abend with an S806 etc... You quickly put it back and everything is back to normal, meanwhile you have egg on your face. That has been the issue since day 1 of of os/360 and it is still valid today. Unless you have a front end to fetch and keep your own numbers you are playing craps with possibility. Personally I would love to know via SMF say a record is cut with totals every so often (user specifiable) that would help a lot. It would be nice to have other information as well (library found in or link list library number). It seems a reasonable thing to do maybe some one has written a SHARE requirement or should submit one. -unsnip-- I know of a LLA exit written by Greg Price that can help with module counts. Perhaps he could be persuaded to release that exit, along with a batch program to fetch/display the counts and reset the counters. I have used this exit and one reporting mechanism in the past and it seems to work very well. It won't display the CPU usage within each module, but it will at least help to idntify those modules that are invoked by non-EXEC statement means. I cannot share it because of various agreements with Greg, but perhaps he can be persuaded to make something of this nature for a very nominal fee, or perhaps free. (I must point out that every scrap of code that I've ever used from Greg has been letter-perfect and completely defect free.) Rick -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
---snip Does it also tell you what program invoked the module? unsnip-- Unfortunately, it does not tell you the identify of the invoker. Given the nature and the mechanics involved, I can understand why. Rick -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
Thanks. I still see this as a useful tool. Joel Wolpert Performance and Capacity Planning consultant WEBSITE: www.perfconsultant.com - Original Message - From: Rick Fochtman rfocht...@ync.net Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@bama.ua.edu Sent: Thursday, August 20, 2009 1:45 PM Subject: Re: Where to find statistics for fetched programs? ---snip Does it also tell you what program invoked the module? unsnip-- Unfortunately, it does not tell you the identify of the invoker. Given the nature and the mechanics involved, I can understand why. Rick -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
Rick Fochtman wrote: I know of a LLA exit written by Greg Price that can help with module counts. Thanks for the kind words, Rick. The LLA exit 1 concept is fairly straightforward (Rob Scott also did it in MXI) but only covers LLA-managed programs. To capture most program usage reliably takes a bit more work. If you are prepared to spend money you could check out the recently GAed Tivoli Asset Discovery for z/OS which happens to employ some of the Dorana technology I helped write way back last millenium. This product is also the logical successor to TLCMz, if I'm not mistaken. If you are not prepared to spend money then the LLA exit approach can still yield interesting data for your site, and the more libraries you LLA-manage the more data you could get by that method. (I used the LLA exit as part of my IMON system monitor thingie to get real-time stats displays in 1995, much as Rob later did with MXI.) Cheers, Greg -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
That is the purpose of the STROBE, GTF, ETC. step. You have to start someplace. The stuff under the covers is the fun part! snip ---snip- Using SMF30 subtype 4 Calculate top 100 cpu consuming steps (Total of TCB/SRB) by program name (EXEC PGM=) over some period of time. Add any appropriate filters and retry. E.G. IEBGENER - IBM code, can't do much about it in the short run, so ignore. Calculate for the top 100 CPU consumed/IO (excp count). Rank the top 100 by CPU/IO. Start at the top of the list and work down using STROBE, GTF, other to determine the cause. Code/test/promote. Repeat ad infinitum, until be benfits gained are exceeded by the effort required. Hope this approach helps. unsnip This sounds good until you dig a little deeper. There's no record in SMF of any subroutines, etc. that are invoked via LINK or LOAD/CALL/DELETE or ATTACH or XCTL. This little gap can lead to a lot of missing formation. BTDTGTSS /snip -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs
The Module Fetch Monitor is still available upon signing of a license agreement for non-warranted programs. Support of any kind is very limited (close to non-existent). Being a non-warranted program, it usually would not be appropriate to open a PMR if there is a problem in this tool. It has been mentioned in Cheryl Watson's newsletter among other places.. Peter Relson z/OS Core Technology Design -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
--snip--- What you said is true but by starting with the data in the SMF you can then drill down. If a program is called via CALL or any of the other methods the CPU will be shown in the SMF for the program in the JCL pgm=. unsnip-- That's true, but if you've got a real PIG of a subroutine that's invoked via LINK, etc. the CPU time will be accounted as part of the main program's CPU time. Not necessarity a good picture, expecially if this subroutine is used by multiple callers. Rick -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
--- On Wed, 8/19/09, Rick Fochtman rfocht...@ync.net wrote: SNIP--- That's true, but if you've got a real PIG of a subroutine that's invoked via LINK, etc. the CPU time will be accounted as part of the main program's CPU time. Not necessarity a good picture, expecially if this subroutine is used by multiple callers. Rick ---end Rick, That is very true. One side issue is that every once in a while there is a program that *NOBODY* knows about and you have done your home work researching where it might be called from. You cannot find any obvious answer so after notifying everyone you delete the module and all of a sudden a lot of programs abend with an S806 etc... You quickly put it back and everything is back to normal, meanwhile you have egg on your face. That has been the issue since day 1 of of os/360 and it is still valid today. Unless you have a front end to fetch and keep your own numbers you are playing craps with possibility. Personally I would love to know via SMF say a record is cut with totals every so often (user specifiable) that would help a lot. It would be nice to have other information as well (library found in or link list library number). It seems a reasonable thing to do maybe some one has written a SHARE requirement or should submit one. Ed -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Where to find statistics for fetched programs?
Hi, I got the question how frequently programmes are fetched, and then how much CPU they consume. This in order to put together kind of a hit parade of big CPU consumers then tackle their tuning. RMF can't help me while Monitor I and Monitor III provide long-term data collection about system workload and resource utilization, while covering all hardware and software components of the system. Monitor II neither while it's about snapshot monitoring. I had a look at SMF too: RecordType4(04) (StepTermination) or Record Type 30 (1E) Common Address Space Work - subtype 4 (Steptotal) is a step in the right direction, but it's on step level. But at first glance, we are rather interested in load module fetch counts CPU consumption independent of step/job. Please route me to what-and-where-to-read. Which utilities or analysis programs can ease the data collection analysis? MXG? Omegamon? Other? I would appreciate you putting me on the right orbit. JV -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
Do you have STROBE or MXG or MICS? SAS? Lizette Hi, I got the question how frequently programmes are fetched, and then how much CPU they consume. This in order to put together kind of a hit parade of big CPU consumers then tackle their tuning. RMF can't help me while Monitor I and Monitor III provide long-term data collection about system workload and resource utilization, while covering all hardware and software components of the system. Monitor II neither while it's about snapshot monitoring. I had a look at SMF too: RecordType4(04) (StepTermination) or Record Type 30 (1E) Common Address Space Work - subtype 4 (Steptotal) is a step in the right direction, but it's on step level. But at first glance, we are rather interested in load module fetch counts CPU consumption independent of step/job. Please route me to what-and-where-to-read. Which utilities or analysis programs can ease the data collection analysis? MXG? Omegamon? Other? -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
I will find out with my (new) customer, Lizette. thanks, JV - Oorspronkelijk bericht - Van : Lizette Koehler [mailto:stars...@mindspring.com] Verzonden : dinsdag , augustus 18, 2009 10:36 PM Aan : IBM-MAIN@bama.ua.edu Onderwerp : Re: Where to find statistics for fetched programs? Do you have STROBE or MXG or MICS? SAS? Lizette Hi, I got the question how frequently programmes are fetched, and then how much CPU they consume. This in order to put together kind of a hit parade of big CPU consumers then tackle their tuning. RMF can't help me while Monitor I and Monitor III provide long-term data collection about system workload and resource utilization, while covering all hardware and software components of the system. Monitor II neither while it's about snapshot monitoring. I had a look at SMF too: RecordType4(04) (StepTermination) or Record Type 30 (1E) Common Address Space Work - subtype 4 (Steptotal) is a step in the right direction, but it's on step level. But at first glance, we are rather interested in load module fetch counts CPU consumption independent of step/job. Please route me to what-and-where-to-read. Which utilities or analysis programs can ease the data collection analysis? MXG? Omegamon? Other? -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
Using SMF30 subtype 4 Calculate top 100 cpu consuming steps (Total of TCB/SRB) by program name (EXEC PGM=) over some period of time. Add any appropriate filters and retry. E.G. IEBGENER - IBM code, can't do much about it in the short run, so ignore. Calculate for the top 100 CPU consumed/IO (excp count). Rank the top 100 by CPU/IO. Start at the top of the list and work down using STROBE, GTF, other to determine the cause. Code/test/promote. Repeat ad infinitum, until be benfits gained are exceeded by the effort required. Hope this approach helps. -Original Message- From: IBM Mainframe Discussion List [mailto:ibm-m...@bama.ua.edu] On Behalf Of Jan Vanbrabant Sent: Tuesday, August 18, 2009 3:26 PM To: IBM-MAIN@bama.ua.edu Subject: Where to find statistics for fetched programs? Hi, I got the question how frequently programmes are fetched, and then how much CPU they consume. This in order to put together kind of a hit parade of big CPU consumers then tackle their tuning. RMF can't help me while Monitor I and Monitor III provide long-term data collection about system workload and resource utilization, while covering all hardware and software components of the system. Monitor II neither while it's about snapshot monitoring. I had a look at SMF too: RecordType4(04) (StepTermination) or Record Type 30 (1E) Common Address Space Work - subtype 4 (Steptotal) is a step in the right direction, but it's on step level. But at first glance, we are rather interested in load module fetch counts CPU consumption independent of step/job. Please route me to what-and-where-to-read. Which utilities or analysis programs can ease the data collection analysis? MXG? Omegamon? Other? I would appreciate you putting me on the right orbit. JV -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
ISTR that Peter Relson had a module fetch monitor tool that he made available on an as is/no warranty basis. He might chime in if he's still paying attention on ibm-main. -- This email might be from the artist formerly known as CC (or not) You be the judge. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
Why do you care about the fetch counts. I would take another approach. Use the SMF type 30 records to determine the jobs that use the most CPU and then see what steps within the job are using the most and tackle their tuning. You can also look at the step termination records as you mentioned below and determine from there what steps use the most CPU and tackle those programs.Strobe is an excellent tool but works at the job level so you still have to determine what jobs to strobe. In the past I had used SAS with MXG to provide a pig report of the top 100 cpu consuming jobs. We used that to determine what jobs and programs to analyze. Also, keep in mind that by analyzing this at the job level you may realize that certain steps can be eliminated or merged by doing things better and smarter. In my opinion just looking at fetch counts and programs is a very narrow view if you are trying to reduce overall mainframe CPU consumption. Just my .02. Joel Wolpert Performance and Capacity Planning consultant WEBSITE: www.perfconsultant.com - Original Message - From: Jan Vanbrabant jan.vanbrab...@telenet.be Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@bama.ua.edu Sent: Tuesday, August 18, 2009 4:25 PM Subject: Where to find statistics for fetched programs? Hi, I got the question how frequently programmes are fetched, and then how much CPU they consume. This in order to put together kind of a hit parade of big CPU consumers then tackle their tuning. RMF can't help me while Monitor I and Monitor III provide long-term data collection about system workload and resource utilization, while covering all hardware and software components of the system. Monitor II neither while it's about snapshot monitoring. I had a look at SMF too: RecordType4(04) (StepTermination) or Record Type 30 (1E) Common Address Space Work - subtype 4 (Steptotal) is a step in the right direction, but it's on step level. But at first glance, we are rather interested in load module fetch counts CPU consumption independent of step/job. Please route me to what-and-where-to-read. Which utilities or analysis programs can ease the data collection analysis? MXG? Omegamon? Other? I would appreciate you putting me on the right orbit. JV -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
---snip- Using SMF30 subtype 4 Calculate top 100 cpu consuming steps (Total of TCB/SRB) by program name (EXEC PGM=) over some period of time. Add any appropriate filters and retry. E.G. IEBGENER - IBM code, can't do much about it in the short run, so ignore. Calculate for the top 100 CPU consumed/IO (excp count). Rank the top 100 by CPU/IO. Start at the top of the list and work down using STROBE, GTF, other to determine the cause. Code/test/promote. Repeat ad infinitum, until be benfits gained are exceeded by the effort required. Hope this approach helps. unsnip This sounds good until you dig a little deeper. There's no record in SMF of any subroutines, etc. that are invoked via LINK or LOAD/CALL/DELETE or ATTACH or XCTL. This little gap can lead to a lot of missing formation. BTDTGTSS Rick -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Where to find statistics for fetched programs?
What you said is true but by starting with the data in the SMF you can then drill down. If a program is called via CALL or any of the other methods the CPU will be shown in the SMF for the program in the JCL pgm=. Joel Wolpert Performance and Capacity Planning consultant WEBSITE: www.perfconsultant.com - Original Message - From: Rick Fochtman rfocht...@ync.net Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@bama.ua.edu Sent: Tuesday, August 18, 2009 9:28 PM Subject: Re: Where to find statistics for fetched programs? ---snip- Using SMF30 subtype 4 Calculate top 100 cpu consuming steps (Total of TCB/SRB) by program name (EXEC PGM=) over some period of time. Add any appropriate filters and retry. E.G. IEBGENER - IBM code, can't do much about it in the short run, so ignore. Calculate for the top 100 CPU consumed/IO (excp count). Rank the top 100 by CPU/IO. Start at the top of the list and work down using STROBE, GTF, other to determine the cause. Code/test/promote. Repeat ad infinitum, until be benfits gained are exceeded by the effort required. Hope this approach helps. unsnip This sounds good until you dig a little deeper. There's no record in SMF of any subroutines, etc. that are invoked via LINK or LOAD/CALL/DELETE or ATTACH or XCTL. This little gap can lead to a lot of missing formation. BTDTGTSS Rick -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
