Using Tivoli Decision Support for z/OS to report on CICS performance

To understand performance data, you must first understand the work CICS® performs at your installation. Analyze the work by its basic building blocks: transactions. Group the transactions into categories of similar resource or user requirements and describe each category’s characteristics. Understand the work that CICS performs for each transaction and the volume of transactions expected during any given period. Tivoli® Decision Support for z/OS® can show you various types of data for the transactions processed by CICS.

A service-level agreement for a CICS user group defines commitments in several areas of quantifiable CICS-related resources and services. CICS service commitments can belong to one of these areas:

The following sections describe certain issues and concerns associated with systems management and how you can use the Tivoli Decision Support for z/OS CICS performance feature.

Monitoring response time

Use the Tivoli Decision Support for z/OS CICS response-time reports to see the CICS application internal response times.

Figure 36 shows the elements of the internal response times.

Figure 36. CICS internal response-time elements
 The response time is the total time from the start to the finish of the transaction' s activity. It is subdivided into suspend time and dispatch time. The dispatch time includes service time.

As described in Decision Support Network Performance Feature Reports, the Network Performance feature generates reports that show the total, end-to-end average response time (operator transit time) for VTAM® applications (for example, a CICS region) by logical unit. The operator transit time consists of the host transit time and the network transit time, which are also shown in the Network Performance feature reports. Using these reports, you can isolate a response-time problem either to the network or to CICS and act on it accordingly. Should the problem be in CICS, you can use the Tivoli Decision Support for z/OS CICS performance feature reports to identify the application causing the response-time degradation.

Monitoring processor and storage use

Poor response time usually indicates inefficient use of either the processor or storage (or both). Tivoli Decision Support-supplied reports can help you isolate a resource as the cause of a CICS performance problem.

If both the Tivoli Decision Support for z/OS CICS performance feature’s statistics component and the Decision Support System Performance feature’s MVS™ component are installed and active, these reports are available for analyzing transaction rates and processor use by CICS region:

Tivoli Decision Support for z/OS produces several reports that can help analyze storage usage. For example, the CICS Dynamic Storage (DSA) Usage report, shows pagepool usage, under the headings 'Pagepool name', 'DSA (bytes)', 'Cushion (bytes)', 'Free storage (bytes)', 'Free storage (pct)', 'Largest free area', 'Getmains', and 'Freemains'.

Figure 37. CICS Dynamic storage (DSA) usage report
                         CICS Dynamic Storage (DSA) Usage
                       MVS ID ='MV28'      CICS ID ='IYCSCTSK'
                        Date:  '2001-01-17' to '2001-01-18'
 
 
                                   Free     Free      Largest
 Pagepool      DSA     Cushion    storage  storage     free
   name      (bytes)   (bytes)    (bytes)   (pct)      area    Getmains  Freemains
 --------  ---------  --------  ---------  -------  ---------  --------  ---------
 CDSA         524288     65536     299008       57     245760      2668       2470
 ECDSA       5242880    131072    1122304       21     868352   1084154    1067000
 ERDSA      11534336    262144    1130496        9     966656       710         16
 ESDSA             0         0          0        0          0         0          0
 EUDSA       2097152         0    2097152      100    1048576     73620      73620
 RDSA         524288     65536     204800       39     122880        40          0
 SDSA         262114     65536     249856       95     249856        12          6
 UDSA         524288     65536     524288      100     262114    301922     301922
 
                 Tivoli Decision Support Report: CICS809

Monitoring volumes and throughput

Because CICS Transaction Server for z/OS uses an MVS subtask to page and because an MVS page-in causes an MVS task to halt execution, the number of page-ins is a performance concern. Page-outs are not a concern because page-outs are scheduled to occur during lulls in CICS processing. If you suspect that a performance problem is related to excessive paging, you can use Tivoli Decision Support for z/OS to report on page-ins, using RMF™ data.

The best indicator of a transaction’s performance is its response. For each transaction ID, the CICS transaction performance detail report (in Figure 38) shows the total transaction count and the average response time. The headings are 'Tran ID', 'Tran count', 'Average resp time (sec)', 'Average CPU time (sec)', 'Prog load reqs (avg)', 'FC calls (avg)', 'Exceptions', 'Program storage bytes (max)', 'Getmains < 16MB (avg)', and 'Getmains > 16MB (avg).'.

Figure 38. CICS transaction performance, detail report
                     CICS Transaction Performance, Detail
                      MVS ID ='MV28'  CICS ID ='IYCSCTSK'
                     Date: '2001-01-17'  to  '2001-01-18'
 
 
                  Avg     Avg   Prog                       Program
                  resp    CPU   load  Prog   FC            storage    Getmains  Getmains
 Tran   Tran     time    time   reqs  loads calls  Excep-  bytes       < 16 MB > 16 MB
  ID    count    (sec)   (sec)  (avg) (avg) (avg)  tions   (max)       (avg)     (avg)
 ------------ -------- -------  ----  ----- -----  ------ ---------    --------  --------
 QUIT    7916    0.085   0.017     0     0    18      0     74344         22         0
 CRTE    1760    4.847   0.004     0     0     0      0    210176          1         0
 AP00    1750    0.184   0.036     0     0     8      0    309800         66         0
 PM94    1369    0.086   0.012     0     0     6      0    130096         24         0
 VCS1     737    0.073   0.008     2     0     7      0     81200         14         0
 PM80     666    1.053   0.155     1     0    62      0    104568        583         0
 CESN     618    8.800   0.001     0     0     0      0     41608          0         0
 SU01     487    0.441   0.062     4     0   126      0    177536         38         0
 ...
 GC11       1    0.341   0.014     1     0     2      0     37048         10         0
 DM08       1    0.028   0.002     0     0     0      0      5040          3         0
     ========                                              =========
        20359                                              309800
 
                      Tivoli Decision Support Report: CICS101

Use this report to start verifying that you are meeting service-level objectives. First, verify that the values for average response time are acceptable. Then check that the transaction counts do not exceed agreed-to limits. If a transaction is not receiving the appropriate level of service, you must determine the cause of the delay.

Combining CICS and DB2 performance data

For each CICS task, CICS generates an LU6.2 unit-of-work ID. DB2® also creates an LU6.2 unit-of-work ID.

Figure 39 shows how DB2 data can be correlated with CICS performance data using the DB2 token (QWHCTOKN) to identify the task.

Figure 39. Correlating a CICS performance-monitoring record with a DB2 accounting record
 You can match the NETNAME and UOWID from the CICS performance-monitoring record, to the DB2 token (QWHCTOKN) from the DB2 accounting record.

If you match the NETUOWPX and NETUOWSX fields in a CICS record to the DB2 token, you can create reports that show the DB2 activity caused by a CICS transaction.

Monitoring exception and incident data

An exception is an event that you should monitor. An exception appears in a report only if it has occurred; reports do not show null counts. A single exception need not be a cause for alarm. An incident is defined as an exception with severity 1, 2, or 3.

The Tivoli Decision Support for z/OS CICS performance feature creates exception records for these incidents and exceptions:

Figure 40 shows an example of an incidents report, giving information on 'Severity', 'Date', 'Time', 'Terminal operator ID', 'User ID', 'Exception ID', and 'Exception description'.

Figure 40. Example of a Tivoli Decision Support CICS incidents report
                                       CICS Incidents
                              DATE: '2001-01-17' to '2001-01-18'
 
                           Terminal
                           operator  User    Exception           Exception
Sev  Date        Time         ID      ID        ID              description
---  ----------  --------  -------- -------- ------------------ ---------------------------
03   2001-01-17  15.42.03  SYSTEM            TRANSACTION_ABEND  CICS TRANSACTION ABEND AZTS
03   2001-01-18  00.00.00  SYSTEM            TRANSACTION_ABEND  CICS TRANSACTION ABEND APCT
03   2001-01-18  17.37.28  SYSTEM            SHORT_OF_STORAGE   CICS SOS IN PAGEPOOL
03   2001-01-18  17.45.03  SYSTEM            SHORT_OF_STORAGE   CICS SOS IN PAGEPOOL
 
                 Tivoli Decision Support report: CICS002

Tivoli Decision Support for z/OS can pass the exceptions to an Information/Management system.

Unit-of-work reporting

In a CICS multiple region operation (MRO) or intersystem communication (ISC) environment, you can trace a transaction as it migrates from one region (or processor complex) to another and back. The data lets you determine the total resource requirements of the combined transaction as a unit of work, without having to separately analyze the component transactions in each region. The ability to combine the component transactions of an MRO or ISC series makes possible precise resource accounting and chargeback, and capacity and performance analysis.

The CICS UOW Response Times report in Figure 41 shows an example of how Tivoli Decision Support for z/OS presents CICS unit- of-work response times. The headings are 'Adjusted UOW start time', 'Tran ID', 'CICS ID', 'Program name', 'UOW tran count', and 'Response time (sec)'.

Figure 41. Tivoli Decision Support for z/OS CICS UOW response times report
                            CICS UOW Response Times                           
                        Time: '09.59.00' to '10.00.00'                      
                               Date: 2001-01-18                         
 
 
          Adjusted
            UOW                                UOW   Response
           start    Tran  CICS      Program   tran     time
            time     ID    ID        name     count   (sec)
          --------  ----  --------  --------  -----  --------
          09.59.25  OP22  CICSPROD  DFHAPRT       2     0.436
                    OP22  CICSPRDC  OEPCPI22
 
          09.59.26  AP63  CICSPRDE  APPM00        2     0.045
                    AP63  CICSPROD  DFHAPRT
 
          09.59.26  ARUS  CICSPROD  DFHAPRT       3     0.158
                    CSM5  CICSPRDB  DFHMIRS
                    ARUS  CICSPRDC  AR49000
 
          09.59.27  CSM5  CICSPRDB  DFHMIRS       4     0.639
                    CSM5  CICSPRDB  DFHMIRS
                    MQ01  CICSPROD  DFHAPRT
                    MQ01  CICSPRDD  CMQ001
 
           ...
 
                  Tivoli Decision Support report: CICS902

Monitoring availability

Users of CICS applications depend on the availability of several types of resources:

In some cases, an application depends on the availability of many resources of the same and of different types, so reporting on availability requires a complex analysis of data from different sources. Tivoli Decision Support for z/OS can help you, because all the data is in one database.

CICS workload activity reporting

CICS records at the end of each transaction:

This is useful when you require only transaction statistics, rather than the detailed information that CMF produces. In many cases, it may be sufficient to process only this data, since RMF records it as part of its SMF type-72 record. Analysis (and even recording) of SMF records from CMF can then be reserved for those circumstances when the detailed data is needed. Use the MVS Performance Management (MVSPM) component of the System Performance feature of Tivoli Decision Support to report on this data.

When running under goal mode in MVS 5.1.0 and later, CICS performance can be reported in workload groups, service classes, and periods. These are a few examples of Tivoli Decision Support reports for CICS in this environment. Figure 42 shows how service classes were served by other service classes. This report is available only when the MVS system is running in goal mode. The headings are 'Workload group', 'Service class', 'Served class', 'No of times served', 'No of transactions', and 'No of times served per transaction'.

Figure 42. Example of an MVS Performance Management served service classes overview report
                         MVSPM Served Service Classes, Overview                         
                         Sysplex: 'SYSPLEX1' System: IP02                         
                         Date: '2001-01-18' Period: 'PRIME'                         
 
 
      Workload Service  Served    No of times         No of    No of times
       group    class   class       served            tx's    served per tx
      -------- -------- -------- ------------  ------------  --------------
      CICS     CICSREGS CICS-1          15227           664            22.9
                        CICS-2           6405           215            29.8
                        CICS-3          24992          1251            20.0
                        CICS-4          87155          1501            58.1
                        CICSTRX         67769          9314             7.3
 
                        Tivoli Decision Support report: MVSPM79

Figure 43 shows the average transaction response time trend and how the various transaction states contribute to it.

Figure 43. Example of an MVS Performance Management response time breakdown, hourly trend report
 The graph shows the average response time, in seconds, over the course of a day. The average response time is shown as a line, and it varies between about 1.5 seconds and nearly 2.5 seconds. The various transaction states are shown as shaded areas stacked above each other, and added together, they represent between a half and two-thirds of the average response time, depending on the time of day. The transaction states shown are Active, Ready, Idle, Lock wait, I/O wait, Conv wait, Distr wait, Syspl wait, Timer wait, Other wait, and Miscellaneous wait. For any time of day, you can read off the average response time, and the average proportion of that time spent in each transaction state state. For example, at 1300 hours, there was an average response time of just over 2 seconds. On average, about 0.5 seconds of that time were spent in an I/O wait. For 0.2 seconds, the transaction was idle. For 0.4 seconds, the transaction was ready. About 0.3 seconds were spent in a number of the other wait states. Adding together all the time spent in the different transaction states gives an execution time of 1.4 seconds, leaving a difference of about 0.6 seconds between the response time and the execution time.

Note that the times shown for the various transaction states are calculated based on transaction state samples, and so are not necessarily a precise record of the time spent in each state. Adding together the time spent in each of the transaction states (the shaded areas on the graph) gives the average execution time, which is lower than the average response time (the line on the graph). The difference between the response time and the execution time is mainly made up of switch time -- for example, the time the transactions spend being routed to another region for processing.

This report is available when the MVS system is running in goal mode and when the subsystem is CICS or IMS™.

Figure 44 shows how much the various transaction states contribute to the average response time. This report is available when the MVS system is running in goal mode and when the subsystem is CICS or IMS. The report gives information on 'Workload group', 'Service class/Period', 'Ph', 'MVS sys ID', and 'Total state', followed by the percentage of response time spent in each of the states listed in Figure 43.

Figure 44. Example of an MVS Performance Management response time breakdown overview report
                                     MVSPM Response Time Breakdown, Overview
                                     Sysplex: 'SYSPLEX1' Subsystem: IP02
                                     Date: '2001-01-18' Period: 'PRIME'
 
 
          Service        MVS  Total Activ Ready Idle   Lock  I/O   Conv Distr Local  Netw Syspl Timer Other  Misc
 Workload  class         sys  state state state state  wait  wait  wait wait  wait   wait wait  wait  wait   wait
   group   /Period    Ph  ID    (%)   (%)   (%)   (%)   (%)   (%)   (%)   (%)   (%)   (%)   (%)   (%)   (%)   (%)
 -------- ---------- --- ---- ----- ----- ----- ----- ----- ----- --     --- ----- ----- ----- ----- ----- ----- -----
 CICS     CICS-1  /1 BTE CA0    6.6   0.0   0.0   0.0   0.0   0.0   6.5   0.0   0.0   0.0   0.0   0.0   0.0   0.0
                         C80   29.4   0.0   0.0   0.0   0.0   0.0  14.7   0.0   0.0   0.0   0.0   0.0  14.6   0.0
                         C90    3.8   0.4   1.3   1.5   0.0   0.2   0.5   0.0   0.0   0.0   0.0   0.0   0.0   0.0
                              ----- ----- ----- ----- ----- ----- -----  --- ----- ----- ----- ----- ----- ----- -----
                            *  13.3   0.1   0.5   0.5   0.0   0.1   7.2   0.0   0.0   0.0   0.0   0.0   4.9   0.0
 
 
                  /1 EXE CA0   16.0   0.1   0.2   0.1   0.0  15.5   0.0   0.0   0.0   0.0   0.0   0.0   0.1   0.0
                         C80   14.9   0.1   0.1   0.1   0.0   3.7   0.0   0.0   0.0   0.0   0.0   0.0  11.0   0.0
                         C90   14.0   1.6   4.5   4.8   0.0   3.2   0.0   0.0   0.0   0.0   0.0   0.0   0.0   0.0
                              ----- ----- ----- ----- ----- -----   --- ----- ----- ----- ----- ----- ----- -----
                            *  14.9   0.6   1.6   1.7   0.0   7.4   0.0   0.0   0.0   0.0   0.0   0.0   3.7   0.0
 
 
 IMS      IMS-1   /1 EXE CA0   20.7   0.4   0.7   0.0   0.0   0.0  19.6   0.0   0.0   0.0   0.0   0.0   0.0   0.0
                         C80    1.1   0.2   0.1   0.7   0.0   0.1   0.0   0.0   0.0   0.0   0.0   0.0   0.0   0.0
                         C90   22.2   5.3  11.9   1.2   0.0   0.2   3.6   0.0   0.0   0.0   0.0   0.0   0.0   0.0
                              ----- ----- ----- ----- ----- -----  ---- ----- ----- ----- ----- ----- ----- -----
                            *  14.7   2.0   4.2   0.6   0.0   0.1   7.8   0.0   0.0   0.0   0.0   0.0   0.0   0.0
 
                          Tivoli Decision Support report: MVSPM73

Related concepts
Tivoli Decision Support for z/OS
CICS Performance Analyzer for z/OS (CICS PA)
[[ Contents Previous Page | Next Page Index ]]