This analysis can cover some or all monitored tasks. The INFORMATION operand of the PRINT-TASK-STATISTICS statement defines the amount of detail desired in the output. Further classification by user IDs (USERID-SET), TSNs (TSN-SET), job name (JOBNAME-SET) and job class (JOBCLASS-SET) is possible.
A set name can be defined for a set of tasks via the CREATE-USERID-SET, CREATE-TSN-SET, CREATE-JOBNAME-SET and CREATE-JOBCLASS-SET statements. Output of these sets is initiated with the PRINT-TASK-STATISTICS statement.
Task statistics analysis
A requirement for task analysis is that the TASK monitoring program was active during SM2 monitoring. The task records in the SM2 output file begin with a START-MP record and end with a STOP-MP record for task statistics.
If several START-MP/STOP-MP pairs occur in an analysis period, several task analyses are output, i.e. each set of task statistics is output separately.
The individual tasks to be analyzed during a monitoring operation are output in the order in which they appear in the SM2 output file. If the analysis period is specified so that it begins or ends between a START-MP and a STOP-MP record, any measurement recorded by the TASK monitoring program whose START-MP record falls within the analysis period is output.
Each monitoring operation is analyzed completely, i.e. from START- through STOP-MEASUREMENT-PROGRAM. Information on the times for START- and STOP-MEASUREMENT-PROGRAM can be obtained using the SM2U1 function SHOW-INFORMATION.
Format
PRINT-TASK-STATISTICS |
INFORMATION = *LOW / *HIGH ,TSN = *ALL / *BY-SET-NAME / *NONE / list-poss(64): <alphanum-name 1..4> ,USERID-SET = *NONE / list-poss(8): <alphanum-name 1..16> ,TSN-SET = *NONE / list-poss(8): <alphanum-name 1..16> ,JOBNAME-SET = *NONE / list-poss(8): <alphanum-name 1..16> ,JOBCLASS-SET = *NONE / list-poss(8): <alphanum-name 1..16> |
Operands
INFORMATION =
Defines the degree of detail for the task statistics. The following data is supplied for each task to be analyzed:
Consecutive task number (INDEX)
This number is supplied by the analysis routine to facilitate task identification. This index is incremented by one each time a task record is found.Identification section
This part serves to identify the task. It contains data such as user ID, task sequence number, job name, job class, etc..Monitoring data
The degree of detail is specified by the following operands:
INFORMATION = *LOW
The monitoring data is output without the data for the monitored devices.
INFORMATION = *HIGH
The monitoring data is output with the data for the monitored devices (ACCESSES and SERVICETIME).
TSN =
Defines the tasks to be analyzed.
TSN = *ALL
All tasks in the SM2 output file are analyzed according to the specified degree of detail.
TSN = *BY-SET-NAME
Specifies exactly the tasks that have been selected by means of a SET name in one of the operands USERID-SET, TSN-SET, JOBNAME-SET or JOBCLASS-SET.
TSN = *NONE
None of the tasks defined by TSN is analyzed. Only those tasks affected by class formation (see the CREATE-USERID-SET, CREATE-TSN-SET, CREATE-JOBNAME-SET, CREATE-JOBCLASS-SET statements) are analyzed if set names are specified for the operands described below.
TSN = list-poss(64): <alphanum-name 1..4>
The tasks defined by the task sequence number (TSN) are analyzed.
USERID-SET =
Selects tasks defined by the user ID for analysis.
USERID-SET = *NONE
None of the tasks defined by the user IS is analyzed.
USERID-SET = list-poss(8): <alphanum-name 1..16>
Set name defined in the CREATE-USERID-SET statement. The set name appears in the class header. Up to 8 such identifiers with 32 user IDs each may be formed. When these limits are exceeded, the surplus elements are ignored. SM2R1 lists the set names and the associated user IDs again on a separate sheet. The output values are total values for all the tasks in a class.
TSN-SET =
Selects tasks defined by means of the task sequence number (TSN) for analysis.
TSN-SET = *NONE
None of the tasks defined by the TSN is analyzed.
TSN-SET = list-poss(8): <alphanum-name 1..16>
Set name defined in the CREATE-TSN-SET statement. The set name appears in the class header. Up to 8 classes with 32 TSNs each may be formed. When these limits are exceeded, the surplus elements are ignored. SM2R1 lists the set names and the associated TSNs again on a separate sheet. The output values are total values for all the tasks in a class.
JOBNAME-SET =
Selects tasks specified by job name.
JOBNAME-SET = *NONE
No tasks specified by job name are selected.
JOBNAME-SET = list-poss(8): <alphanum-name 1..16>
Set name defined by the CREATE-JOBNAME-SET statement. The set names appears as the header for this class. Up to 8 such classes with 32 associated job names can be formed. SM2R1 lists the set names with associated job names again on a separate sheet. The output values are total values for all the tasks in a class.
JOBCLASS-SET =
Selects tasks specified by job name.
JOBCLASS-SET = *NONE
No tasks specified by job name are selected.
JOBCLASS-SET = list-poss(8): <alphanum-name 1..16>
Set name defined by the CREATE-JOBCLASS-SET statement. The set name appears as the header of this class. Up to 8 such classes with 32 associated job classes can be formed. SM2R1 lists the set names with the associated job classes on a separate sheet. The output values are total values for all tasks belonging to the class.
If the user specifies the set name, all tasks which do not belong to the above classes are combined to form a residual class. If the residual class is devoid of tasks, there is no output.
Example
//PRINT-TASK-STATISTICS TSN=(BCAM,PGE)
This statement supplies an analysis output for the tasks with the TSNs BCAM and PGE.
Task analysis supplies the following summaries by default.
To facilitate understanding, the following terms are defined:
TASK START TIME: time at which task monitoring is started. For tasks which did not exist when monitoring was started, this is the LOGON time. Only for system tasks is this the task creation time.
TASK STOP TIME: time at which monitoring of the task terminates. For tasks which are still running at monitoring termination time, this time is the task statistics STOP time (see the STOP-MP record). For tasks which were terminated before monitoring termination, this time is the task termination time (LOGOFF or task destruction).
TASK MONITORING CYCLE: difference between task stop time and task start time.
VOLUNTARY ACTIVE WAIT STATE: when a task is active, SM2 interprets the following times as voluntary wait times:
time spent in queue 2
time spent in queue 4 for boursing
time spent in queue 4 for inter-task communication (ITC)
time spent in queue 4 for PASS and VPASS
VOLUNTARY INACTIVE WAIT STATE: When a task is inactive, SM2 interprets the following times as voluntary wait times:
time spent in queue 12 for boursing (this time is displayed separately)
time spent in queue 13 (PASS, VPASS, ITC)
Summaries supplied by task statistics by default
List sorted in descending order of the 20 tasks which used up most CPU time in the task monitoring cycle. For each task the consecutive task number (INDEX), the user ID (USERID), the task sequence number (TSN), the task category (CATEGORY), the CPU time (CPU-TIME) used in seconds, the number of I/O operations (#I/O'S), the name of the program with the highest CPU utilization (PROGRAM-NAME) and its CPU utilization (PGM-CPU) are supplied.
List sorted in descending order of the 20 tasks which performed most I/O operations during the task monitoring cycle (without paging I/Os). The same data is output for each task as in the preceding summary, referring this time to the number of I/O operations.
List sorted in descending order of the 20 tasks which occupy the maximum user address space. The tasks are sorted according to the maximum amount of class 5 and class 6 memory.
The monitored tasks are grouped according to category and the category data is output. Totals are supplied on a system basis (SUM) and per category. The degree of detail is the same as for LEVEL=HIGH or LEVEL=LOW without the identification part. It should be noted that the PGE task for device-specific ACCESSES and SERVICETIME values (data supplied per task) is not included in this summary.
Not more than the first 30 categories found in the SM2 output file (including the pseudocategory SUM) are output for sum and class; they are output in the following order: SUM, standard categories SYS, DIALOG, BATCH and TP, remaining categories (in alphabetical order). If more than 30 categories occur in a monitoring cycle, the surplus categories are ignored.
This does not affect task category output. The category to which the task was assigned at termination time is always output.
Output data per task
Identification part
The identification part comprises the following data:
USERID
User identification
Blanks are output if the user ID cannot be determined (e.g. for system tasks).TSN
Task sequence number
JOBNAME
Job name (blanks are entered for system tasks)
JOBCLASS
Name of job class (blanks are entered for system tasks)
CATEGORY
Task category at task stop time
START-DATE
Date at task start time
START-TIME
Time of day at task start time
END-DATE
Date at task stop time
END-TIME
Time of day at task stop time
Data supplied for LEVEL=LOW
ELAPSED TIME(S)
Same as task monitoring cycle.
TOTAL CPU-TIME(S)
Accumulated CPU time (TU and TPR shares of CPU time) in seconds since task initiation.
TOTAL # IO'S
Total number of input/output operations (except for paging) since task initiation (number of EXCP calls).
# IO'S
Total number of input/output operations (except for paging) in the task monitoring cycle (number of EXCP calls).
# SVC'S IN TU STATE
Number of SVC calls from the TU state in the task monitoring cycle.
# SVC'S IN TPR STATE
Number of SVC calls from the TPR state in the task monitoring cycle.
This number does not contain the FUNCTION DETACH calls.
For the four following items refer to “page fault” in the glossary.
# PAGE FAULTS
Number of page fault interrupts in the task monitoring cycle. “Real” page faults (access to virtually unassigned pages) are not included in this number.
# PAGE READS
Number of pages read from background memory during the task monitoring cycle.
# PAGE RECLAIMS
Number of page fault interrupts in the task monitoring cycle for which the addressed page was still in main memory.
# 1ST PAGE ACCESSES
Number of page fault interrupts for first access to a page in the task monitoring cycle.
WSET (PPC) IN PAGES
Average working set value (PPC) during the task monitoring cycle.
# ESA PAGES
Maximum occupancy of DATA SPACE pages (only for the task which has created the data space).
# CLASS 5 + 6 PAGES
Maximum occupancy of user address space (total amount of class 5 and class 6 memory, including any common memory pools present).
# CPU ALLOCATIONS
Number of requests to the CPU(s) during the task monitoring cycle. This number does not contain the requests for which the task retains the CPU after interrupts. (all exits from queue 0 to other queues are counted.)
CPU-TIME (S)
CPU time (TU and TPR shares of the CPU time) in the task monitoring cycle in seconds.
# CPU WAITS
Number of wait states for CPU allocation (queue Q1)
DURATION (S)
Total time in seconds spent waiting for CPU allocation.
This also includes periods in which the CPU is active for other tasks in the SIH state or is occupied with management tasks.# ACTIVE WAITS
Number of active wait states in the TASK-MESSPERIODE (queues Q2 and Q4 except for time for I/Os)
DURATION (S)
Total dwell time (in seconds) during voluntary active wait states in the task monitoring cycle.
# DISK IO WAITS
Number of wait states for I/O operations on disk devices in the task monitoring cycle (queues Q4 and Q12).
DURATION (S)
Total dwell time (in seconds) for I/O operations to disk devices in the task monitoring cycle.
# NON DISK IO WAITS
Number of wait states for I/O operations during the wait states in the task monitoring cycle (queues Q4 and Q12).
DURATION (S)
Total dwell time (in seconds) during voluntary inactive wait states in the task monitoring cycle except for boursing.
# INACTIVE WAITS
Number of voluntary inactive wait states in the task monitoring cycle except for boursing (queues Q10, Q11, Q13).
DURATION (S)
Total dwell time (in seconds) during voluntary inactive wait states in the task monitoring cycle except for boursing.
# ADMISSIONS
Number of admissions in the task monitoring cycle (sum of queues Q5 and Q6).
DURATION (S)
Wait time for admission in seconds (sum of queues Q5 and Q6).
# BOURSE LONG WAITS
Number of voluntary inactive wait states in the task monitoring cycle due to boursing (queue 12).
DURATION (S)
Total dwell time (in seconds) during voluntary inactive wait states in the task monitoring cycle due to boursing.
SERVICE UNITS
Service units which have been included in the task monitoring cycle.
CPU SERVICE UNITS
CPU service units which have been included in the task monitoring cycle.
IO SERVICE UNITS
I/O service units which have been included in the task monitoring cycle.
MEMORY SERVICE UNITS
Memory service units which have been included in the task monitoring cycle.
For the following four items, the name of the most frequently used program is output. The program name is specified as follows:
the file name, in the case of linked programs (START-EXECUTABLE-PROGRAM FROM-FILE= <filename>). If the file name including catalog ID and user ID is longer than 20 characters, the catalog and user IDs are omitted from the file name and only the first 20 characters are output.
the module name, in the case of programs started via DLL (START-EXECUTABLE-PROGRAM FROM-FILE=*MODULE(LIBRARY=*STD,ELEMENT=...))
EAM OMF in the case of programs started from the EAM library (START-EXECUTABLE-PROGRAM FROM-FILE=*MODULE(LIBRARY=*OMF,ELEMENT=...))
*NONE, if no program was executed within the task during the monitoring cycle.
HIGHEST CPU CONSUMER
The program with the highest CPU consumption (in seconds).
HIGHEST IO CONSUMER
The program with the most I/O operations.
HIGHEST S-U CONSUMER
The program with the highest service unit consumption.
HIGHEST WS CONSUMER
The program with the highest working set consumption.
HIGHEST PRIORITY
The highest priority assigned for the task.
LOWEST PRIORITY
The lowest priority assigned for the task.
# INPUT MESSAGES/SEC
Number of input messages per second for this task. Contrary to the RESPONSETIME monitoring program, messages between applications within a host are also included.
AVG INP.-LGTH (BYTES)
Average input length in bytes.
# OUTPUT MESSAGES/SEC
Number of output messages per second from this task. Contrary to the RESPONSETIME monitoring program, messages between applications within a host are also included.
AVG OUTP.-LGTH (BYTES)
Average output length in bytes.
PAMPAGES PER DISK IO
Average number of PAM pages per I/O operation on disk devices during the task monitoring cycle.
KB PER NON DISK IO
Average amount of data transported in Kb per I/O output operation on other devices during the task monitoring cycle.
Values supplied if LEVEL=*HIGH
In this case the same values as for LEVEL=*LOW are supplied together with those described below.
Two values, subdivided on DVS input/output) and paging (DVS or PAGING) are supplied for each monitored device specified using a mnemonic device name. The mnemonic device name and the pubset ID are output.
This data is also supplied, broken down as to PAGING and DMS values, as total values for all devices.
ACCESSES PER SEC
Number of accesses (EXCP calls) per second and per device
ACCESSES TOTAL
Number of accesses (EXCP calls) per device
SERVICETIME HW (MS)
Average hardware service time per device.
SERVICETIME SW (MS)
Average software service time per device. (including hardware service time)
In the case of paging I/Os, only input/output operations are assigned to the task; these operations relate to the paging error interrupt caused by the task, i.e. in the case of read chains, these I/O operations are assigned to the relevant task (the counter is incremented) and the hardware time is included in the calculations.
For parallel access volumes (PAV), the devices specified for the monitoring program definition are output. The data of the basic devices refers to the basic device and the assigned alias devices.
Write-only inputs/outputs are assigned to the PGE task (see “assignment of paging activities to the initiating task” in the glossary).
Data supplied in addition for classes (user ID, TSN, categories, job name, job class).
# TASKS
Number of tasks in this class.
AVERAGE MPL
Average MPL (Multi Programming Level).
This value is computed from the elapsed time for all tasks in the class, divided by the time from START-MP to STOP-MP. Inaccuracies can occur as a result of the fact that ELAPSED TIME and START/STOP-MP time are measured in different units.
The number of device accesses per second is computed from the total number of IOs per device, divided by the time from START-MEASUREMENT-PROGRAM to STOP-MEASUREMENT-PROGRAM or from task creation at time of task terminating.
The monitoring data # ADMISSIONS and DURATION (S) has only values different from zero if the PCS was in operation for part of the time or the entire time. For the interpretation of the data see the “PCS” manual [12 (Related publications)].
For SPOOL tasks the accumulated CPU time and the number of I/O operations (TOTAL CPU-TIME(S) or TOTAL # IO'S) may be incorrect because the system counters monitored by SM2 are reset to zero by SPOOL each time a job terminates.
If a SPOOL task is sensed at STOP-MEASUREMENT-PROGRAM time and a SPOOL job is in progress at this time, the TSN and user ID of the initiator are output.