Upload
ibm-ims
View
362
Download
6
Tags:
Embed Size (px)
Citation preview
Trademarks and disclaimers
The following are trademarks of the International Business Machines Corporation in the United States, other countries, or both.
For a complete list of IBM Trademarks, see http://www.ibm.com/legal/us/en/copytrade.shtml
Not all common law marks used by IBM are listed on this page. Failure of a mark to appear does not mean that IBM does not use the mark nor
does it mean that the product is not actively marketed or is not significant within its relevant market. Those trademarks followed by ® are
registered trademarks of IBM in the United States; all others are trademarks or common law marks of IBM in the United States.
Any opinions, findings, conclusions or recommendations presented in this material are strictly those of the speaker and do not necessarily
reflect the views of the Arizona Strategic Enterprise Technology (ASET), Arizona Department of Administration (ADOA) or the State of Arizona.
Organic New/changed application
Occasional Spikes
Capacity limits or SLA
Observed behavior
Usage and Growth
Organic New/changed application
Occasional SpikesData Growth:• Old call patterns• I/O bottlenecks
Usage and Growth
Organic New/changed application
Occasional Spikes
Installation successBut…• Data grows with use• Widespread adoption
Classic TIP-TOP
Usage and Growth
Organic New/changed application
Occasional Spikes
Was everything tested with this change?
Usage and Growth
Organic New/changed application
Occasional Spikes
Most likely causes:• Ad-hoc programs…• …in PRODUCTION
((Yikes!))
Secondary cause:• Competition for resources
See Next Chart!!!!
Usage and Growth
“…If we tune the environment, we can reduce MIPS being used by the application…”
“…Performance tuning is easily resolved by finding the spikes and correcting for those spikes…”
“…We’re getting performance problems in ______. We did not change anything…”
Better approach:
» Tune the application
» Put the “spikes” in context
˃ New app(s)?
˃ Change in app(s)?
˃ Change in environment?
˃ Continued growth?
» Consistent measurement –
How do you know there is poor
performance?
» Usually do not happen when applications are…
˃ Unit tested
˃ System tested or shakedown tested
˃ Integration tested
» But when they happen, DBAs ask…
˃ Is this a one-off situation?
˃ Or is this a new “steady state”?
+ New workload
+ Existing work changes
+ Programmer/DBA misstep
» General performance goals…
˃ Increase throughput
˃ Reduce MIPS
» As for performance tuning in general…
˃ Cross-verify
˃ 1,000 paper cuts
˃ Service level agreement
» Type 42 – SMS I/O Records
» Type 70:79 – RMF Records
» Type 100:102 – DB2 Records
» Type 110 – CICS Records
19
28
* Subsystem Type IMS - IMS trans response goals
Classification:
Default service class is IMS
There is no default report class.
Qualifier Qualifier Starting Service Report
# type name position Class Class
- ---------- -------------- --------- -------- --------
1 SI IMSP* IMS BATIMSP
2 . TNG . UNITRANS IMSHOT IMSPTRN1
2 . TNG . MVDHOT MVDHOT IMSPTRN2
2 . TN . * MVDMED IMSPTRN3
1 SI IMST* IMST BATIMST
2 . TNG . UNITRANS IMSHOT IMSTTRN1
2 . TN . * IMST IMSTTRN3
1 SI IMSE* IMST BATIMSE
1 SI IMSF* IMST BATIMSF
1 SI IMSG* IMST BATIMSG
29
* Transaction Name Group MVDHOT - mvd high priority trans
Qualifier
name Description
--------- --------------------------------
EVFEE fee entry
ICEZALT saz vehicle renewal
MDADR modify address
MDAPP modify application
MDEDL saz drivers application
MDIBMN saz driver motor veh record
MVPTL vehicle title print
MVPTL2 vehicle title print part2
MVRAT vehicle reg and transfer
MVRBM2 renew by mail
MVREG modify registration
MVTRP temp reg permits
QDSEE driver inquiry
QVREG vehicle inquiry
30
* Subsystem Type IMS - IMS trans response goals
Classification:
Default service class is IMS
There is no default report class.
Qualifier Qualifier Starting Service Report
# type name position Class Class
- ---------- -------------- --------- -------- --------
1 SI IMSP* IMS BATIMSP
2 . TNG . UNITRANS IMSHOT IMSPTRN1
2 . TNG . MVDHOT MVDHOT IMSPTRN2
2 . TN . * MVDMED IMSPTRN3
1 SI IMST* IMST BATIMST
2 . TNG . UNITRANS IMSHOT IMSTTRN1
2 . TN . * IMST IMSTTRN3
1 SI IMSE* IMST BATIMSE
1 SI IMSF* IMST BATIMSF
1 SI IMSG* IMST BATIMSG
31
* Service Class IMS - Prod IMS response time goal
Base goal:
CPU Critical flag: NO
# Duration Imp Goal description
- --------- - ---------------------------------------
1 1 99% complete within 00:00:00.200
* Service Class IMSHOT - Prod IMS resp time < .1 sec
Base goal:
CPU Critical flag: YES
# Duration Imp Goal description
- --------- - --------------------------------------
1 1 99% complete within 00:00:00.100
* Service Class MVDHOT - Prod IMS resp < .5 sec
Base goal:
CPU Critical flag: YES
# Duration Imp Goal description
- --------- - ---------------------------------------
1 1 99% complete within 00:00:00.500
* Service Class MVDMED - Prod IMS resp < 1.0 sec
Base goal:
CPU Critical flag: YES
# Duration Imp Goal description
- --------- - ---------------------------------------
1 1 99% complete within 00:00:01.000
32
* Service Class IMS - Prod IMS response time goal
Base goal:
CPU Critical flag: NO
# Duration Imp Goal description
- --------- - ---------------------------------------
1 1 99% complete within 00:00:00.200
* Service Class IMSHOT - Prod IMS resp time < .1 sec
Base goal:
CPU Critical flag: YES
# Duration Imp Goal description
- --------- - --------------------------------------
1 1 99% complete within 00:00:00.100
* Service Class MVDHOT - Prod IMS resp < .5 sec
Base goal:
CPU Critical flag: YES
# Duration Imp Goal description
- --------- - ---------------------------------------
1 1 99% complete within 00:00:00.500
* Service Class MVDMED - Prod IMS resp < 1.0 sec
Base goal:
CPU Critical flag: YES
# Duration Imp Goal description
- --------- - ---------------------------------------
1 1 99% complete within 00:00:01.000
41
IMSPSTC1 IMP=1 Velocity=70IMSPROD – IMS Control Region
IMSPSTC2 IMP=2 Velocity=92IMSPDLI – IMS DLI (I/O) Address space
IMSPSTC3 IMP=2 Velocity=60IMSPM* - Dependent IMS DC Control Regions
SYSSTCIMSPDBRC – DBRC Address space
http://www.redbooks.ibm.com/redbooks/pdf
http://www-03.ibm.com/systems/z/os/zos/features/rmf/tools/rmftools.html
//SMFDMP1 EXEC PGM=IFASMFDP
//DUMPIN DD DISP=SHR,
// DSN=<Your SMF file goes here>
//RMF DD DSN=<Your SMF Extract goes here>,
// DISP=(,CATLG,DELETE),UNIT=SYSDA,
// SPACE=(CYL,(500,50),RLSE)
//SYSPRINT DD SYSOUT=*
//SYSIN DD *
INDD(DUMPIN,OPTIONS(DUMP))
OUTDD(RMF,TYPE(70:79))
DATE(2013024,2013024)
START(0958)
END(1131)
//SMFSORT EXEC PGM=SORT,REGION=0M,
// PARM='EQUALS,HIPRMAX=10,SIZE=8192K'
//SORTLIB DD DSN=SYS1.SORTLIB,DISP=SHR
//SYSOUT DD SYSOUT=*
//SORTWK01 DD UNIT=SYSDA,SPACE=(CYL,(50,50))
//SORTWK02 DD UNIT=SYSDA,SPACE=(CYL,(50,50))
//SORTWK03 DD UNIT=SYSDA,SPACE=(CYL,(50,50))
//SORTWK04 DD UNIT=SYSDA,SPACE=(CYL,(50,50))
//SORTIN DD DCB=BUFNO=40,DISP=SHR,
// DSN=<Your SMF Extract goes here>
//SORTOUT DD DSN=&&RMFSRT,
// SPACE=(TRK,(150,150),RLSE),DISP=(NEW,CATLG),
// DCB=*.SORTIN,UNIT=SYSDA
//SYSIN DD *
SORT FIELDS=(11,4,CH,A,7,4,CH,A),EQUALS
MODS E15=(ERBPPE15,500,,N),E35=(ERBPPE35,500,,N)
RECORD TYPE=V,LENGTH=(,,,18,400)
END
/*
//RMFPP EXEC
PGM=ERBRMFPP,REGION=0M,COND=(0,NE),TIME=60
//STEPLIB DD DSN=SYS1.SERBLINK,DISP=SHR
//SYSOUT DD SYSOUT=*
//SYSPRINT DD SYSOUT=*
//MFPMSGDS DD SYSOUT=*
//PPXSRPTS DD SYSOUT=*
//PPRPTS DD SYSOUT=*
//MFPINPUT DD DISP=SHR,DSN=&&RMFSRT
//SYSIN DD *
SYSOUT(A)
REPORTS(ALL)
SYSRPTS(CF,WLMGL(SCPER,WGROUP,RCLASS,RCPER,POLICY))
/*
» Call pattern and the Unit-of-work (UOW)˃ Were all programs changed reviewed for any update intend?
˃ Do the issue a HOLD for each call as it walks through the database?
˃ Does it take periodic checkpoints to release locks so that other programs are allowed to do updating?
˃ Were other batch routines which run at the same time analyzed as well for HOLDs and lock releases?
» PSB analysis˃ Is there a need for committed reads (concurrent or programmatic relationships) or can the
program exploit the use of dirty reads? (using the ONLY option on the PROCOPT)
˃ Does each of the updaters use appropriate PROCOPTS for the activity performed? (e.g. Using a PROCOPT=A when a PROCOPT=GID would be sufficient)
» Database level˃ Have all databases been REORGed which are being used in the batch routines in conflict? IMS
(just like DB2) does not manage deleted storage for reuse. Secondly, on highly volatile databases, chain-chasing can cause a database to get incrementally slower over time. To counter this, on-line REORGs are a beautiful solution.
˃ Is there a way or a reporting process in place to adequately watch for organic growth?
˃ Have the buffers been carefully scrutinized lately for appropriateness due to any growth? When was the last time the DFSVSAMP buffers were adjusted due to database growth?
» System level˃ Very little can be done here except in the aggregate using workload manager (WLM)
59
» Jeffy's Rule of Mainframe Workload:˃ Given a fixed set of workload on a given machine, as throughput of work
increases (same work performed in less time) MIPS will increase
» Application of this rule: If you want to decrease MIPS you must either
1. Rewrite the workload
2. Detune the environment
3. Underlying DBMS, z/OS, DASD microcode upgrades
4. Some type of system capping
» Do NOT tie MIPS usage to performance
» In the aggregate what are the desirable goals: ˃ Increase throughput
˃ Reduce MIPS
60