CICS Performance And Tuning 101

CICS Performance and Tuning 101Ezriel Gross - Circle Software IncorporatedTuesday, March 8, 2016

Agenda Introduction to tuning and reasons to tune– Application versus systems Tuning methodologyAnatomy of response timeData collection and reporting facilitiesGetting started: monitoring, DFH0STAT, end-of-day (EOD) statisticsExamples of resources to tuneSummary2CICS Performance and Tuning 101

Why tune? Poor response times– Application versus system– Network– DASD Increased workload– Consolidations– Increased volumes Hardware considerations– Postpone upgrades Application costsLearning experience3CICS Performance and Tuning 101

When is tuning usually performed? When problems or issues occur Why not tune regularly?–––––––4Lack of resources due to budget cuts, staff reductionsLack of knowledgeLack of interest – application or file tuningDependence on outside parties“If it ain’t broke don’t fix it” attitudeThird party packages and applicationsROI cannot be determined in advanceCICS Performance and Tuning 101

Rules before starting Tuning is a “top-down” activityMake changes to address major constraintsOne major change at a timeSome changes require iteration to find the right valuesChange should be done gradually and monitoredTuning will not always be effectiveDo not tune for the sake of tuningHave a fallback position5CICS Performance and Tuning 101

Methodology6CICS Performance and Tuning 101

MethodologyGeneral performance tuning guidelines: Observe–––Understand your startup procedure and workloadSet realistic objectivesDevelop a baseline to which you can compare CPU Utilization: overall and CICSNumber of tasks per day or hour: peak and averageResponse timesMeasure–––Identify areas to tuneDetermine measurement timelineSelect tools to be used 7IBM supplied: DFH0STAT, EOD Statistics, CICS tables/RDO information, LISTCAT etc.Third party monitors and toolsCICS Performance and Tuning 101

Methodology Analyze–– Review outputsIdentify tuning opportunitiesReact–Make appropriate changes 8Use test or quality environments firstWord to the wise– Make major changes one at a time– Follow installation standards: change managementEnsure backup or fallback plan is readyCICS Performance and Tuning 101

Methodology Verify––Review results from changesMake appropriate changes, as required – Some tuning may require several iterations: for example LSR pool tuningGo back to the Measure step until changes are meeting your objectivesImplement–9Move to production and go back to the Observe stepCICS Performance and Tuning 101

MethodologyEstablishbaselineIdentify performanceconstraintMonitor the systemusing toolsObjectivemet?YESDetermine requiredchangeNOMake change andimplementDocument changeand continue normalmonitoring10CICS Performance and Tuning 101

no ing CICS or your applications is not enough 11CICS Performance and Tuning 101

Performance opportunities 12Response time problemsProcessor overloadedCPU problems and costsProvision for increased workloadsAvailability and reliability problemsLack of certain types of CICS resourcesCapacity planning base lineRealization of new technologiesCICS Performance and Tuning 101

Response times System response– Allocation of resources– Processor speed– Design of application code Network response– Transmission through the network DASD response– Caching and buffering to reduce or eliminate Increased workload– Can cause failures in otherwise stable environment13CICS Performance and Tuning 101

Response times14CICS Performance and Tuning 101

CICS tasks and programs A task is an instance of a transaction started by a user. When a user types in data and presses Enter or aFunction key, CICS begins a Task and loads thenecessary programs. Tasks run concurrently. Therefore, a user can runmultiple instances of the same transactionsimultaneously. CICS multitasks giving fast response times. CICS runs each task, briefly giving CPU to each one.15CICS Performance and Tuning 101

EXEC interface 16CICS programs look likebatch with the insertion ofExecute CICS commands.The CICS commands areused to request Services.APPLICATIONSYSTEMDFHETCDFHEIFCCICS commands must betranslated into COBOLprior or during programcompilation.WAIT?DFHETSCICS Performance and Tuning 101

Response times17CICS Performance and Tuning 101

Anatomy of response time Response time consists of two elements:1.2.Suspend time: the time a task is not executing (waiting).Dispatch time: the time that CICS thinks the task is executing. Thistime is further divided into:A. CPU time: the time the task is executing on CPU.B. Wait time: the time the CPU has been taken away from the taskwithout the knowledge of CICS. CPU to dispatch ratio:– Ratio (CPU time/dispatch time) * 100– Objective is 80% or higher18CICS Performance and Tuning 101

Types of aits19CICS Performance and Tuning 101

Resource load20CICS Performance and Tuning 101

Instrumentation data and measurement tools21CICS Performance and Tuning 101

CICS monitoring facility (CMF) CMF collects data about all transactions in CICS Records are written to SMF for later offline processing CMF collects four classes of data: exception, identity, performance andtransaction resource CMF can produce a significant volume of data CICS compresses the data by default To exclude monitoring data fields, use a monitoring control table (MCT) To process output use:– Tivoli Decision Support for z/OS– CICS Supplied sample program DFH MOLS– CICS Performance Analyzer (CICS PA)22CICS Performance and Tuning 101

CMF data types Exception Class– Information about resource shortages encountered Queuing for file strings Wait for Temporary Storage buffers– Highlights problems in CICS system operation– Identifies system constraints that affect performance– One exception record written for each condition that occurs Identity Class Provides enhanced audit information Captures identity propagation data from a client system across anetwork for eligible transactions23CICS Performance and Tuning 101

CMF data types Performance Class––––Provides detailed transaction informationProcessor and elapsed timeTime spent waiting for I/O, etc One record per transaction Transaction Resource Class– Additional transaction level information about individual resourcesaccessed by a transaction– Items such DPLs, file and temporary storage queues– One transaction resource record per transaction monitored– Record cut only if transaction accesses at least one resource beingmonitored24CICS Performance and Tuning 101

DFH MOLS Sample program supplied with CICS that can process and printSMF records produced by CMF Can also produce output data sets from compressed records foruse by other analysis products Sorts and prints monitoring data based on parameters passed Can select or exclude data for printing based on applid, userid,tranid, task number, etc. SMF data set must be unloaded prior to using MCT is key to determining how much data is collected25CICS Performance and Tuning 101

CICS Monitoring Control Facility (CEMN)26CICS Performance and Tuning 101

CICS Monitoring Control Facility (CEMN)27CICS Performance and Tuning 101

CICS Monitoring Control Facility (CEMN)28CICS Performance and Tuning 101

CICS Monitoring Control Facility (CEMN)29CICS Performance and Tuning 101

CICS Monitoring Control Facility (CEMN)30CICS Performance and Tuning 101

CICS statistics Statistics domain collects a variety of data Written to the SMF data set Provides information about resources and domains– Counts and wait times for resource requests– Processor and storage use 31Certain statistics counters can be reset when records cutInterval recording can be set on/off using STATRCD (SIT)Records can be processed by DFHSTUP or DFH0STATRecords can be processed by Tivoli Decision Support for z/OS or CICSPerformance Analyzer (CICS PA)CICS Performance and Tuning 101

When does CICS collect statistics? Interval statistics– At intervals set: default every hour– Requires STARCD ON in SIT– Can be turned on using SET command End-of-day statistics– When CICS shuts down either normal or immediate– At midnight (by default) in 24/7 operations Requested statistics––––32EXEC CICS Perform statistics recordEXEC CICS Set statistics RECORDNOWCEMT Perform statisticsCan be issued with any combination of resourcesCICS Performance and Tuning 101

When does CICS collect statistics? Requested Reset statistics––––– EXEC CICS Perform statistics record RESTNOWEXEC CICS Set statistics RECORDNOW RESETNOWCEMT Perform statistics all RESTNOWDiffers from Request Statistics as counters are resetCauses loss of data since the last statistics intervalUnsolicited statistics– Collected for resources allocated or de-allocated– Written to SMF before resource is deleted– Produced for resources such as, Atom Feeds, Autoinstalled Terminals, Files,DB2, FEPI, IPCONN, etc.33CICS Performance and Tuning 101

DFH0STAT RDO Group DFH STAT contains required definitions Transaction identifier is STAT COBOL sample provided in source and load format Good example of the use of EXEC CICS Collect Statistics Output information includes:– 34Dispatcher, Storage, Loader, etc.Spool YES must be specified in the SITCICS Performance and Tuning 101

DFH0STAT35CICS Performance and Tuning 101

DFH0STAT36CICS Performance and Tuning 101

DFH0STAT37CICS Performance and Tuning 101

DFH0STAT38CICS Performance and Tuning 101

DFH0STAT39CICS Performance and Tuning 101

DFH0STAT40CICS Performance and Tuning 101

What resources to tune41CICS Performance and Tuning 101

What resources to tune42CICS Performance and Tuning 101

What resources to tune43CICS Performance and Tuning 101

What resources to tune44CICS Performance and Tuning 101

Understanding serialization45CICS Performance and Tuning 101

Making a change46CICS Performance and Tuning 101

Making a change47CICS Performance and Tuning 101

Maximum Tasks (MXT)48CICS Performance and Tuning 101

Maximum Tasks (MXT)49CICS Performance and Tuning 101

Maximum Tasks (MXT)50CICS Performance and Tuning 101

Transaction Class (TCLASS)51CICS Performance and Tuning 101

Transaction Class (TCLASS)52CICS Performance and Tuning 101

Transaction Class (TCLASS)53CICS Performance and Tuning 101

Task Priority & PRTYAGE (SIT)54CICS Performance and Tuning 101

Task Priority & PRTYAGE (SIT)55CICS Performance and Tuning 101

Task Priority & PRTYAGE (SIT)56CICS Performance and Tuning 101

Task Priority & PRTYAGE (SIT)57CICS Performance and Tuning 101

Region Exit Interval (ICV)58CICS Performance and Tuning 101

Interval Runa ay (ICVR)59CICS Performance and Tuning 101

Interval Runa ay (ICVR)60CICS Performance and Tuning 101

Multi-Tasking (DTIMEOUT)61CICS Performance and Tuning 101

Deadlock Timeout (DTIMEOUT)62CICS Performance and Tuning 101

DISPATCHER STATISTICS63CICS Performance and Tuning 101

Systems of Record can be complex systems built on decades ofcontinuous & incremental developmentChange?Modernize?64Fix problems?Where do I start?Optimize?Re-use?CICS Performance and Tuning 101

CICS Operational InsightTry it out athttps://cicsoi.mybluemix.net/Have your say on new insights gInstrumented65AnalyticsIntegratedInformationCICS Performance and Tuning 101

Three steps to CICS Operational EfficiencyOptimize applications, systems, and processes to achieve more with lessDeveloperSysprogIdentify hotspottransactionsCICS PAIdentify hotspotprogramsCICS IACode changes, ifneededRDzCollect and loadperformance dataCreate modelDev/Test regionsCICS DAClone Dev/TestregionsAdd updatedresources tochange packageWSIMCICS PACICS CMPerformancetestDeploy newcode andchangepackageStep 1Step 2Step 3Threadsafe analysis andimplementation to reduceCPU utilization by up to 20%Workload Management tocope with mobile scalabilityand availability needsRegion Consolidation toreduce management overheadand CPU utilization (up to10%)66CICS Performance and Tuning 101