This database file contains information about job, task, and thread wait conditions.
At least one record will be written for each job, task, or thread that consumed CPU during the interval (multiple records are possible especially during service activities).
The purpose of this
file is to account for the time a job (this means a task, primary thread,
or secondary thread) spends waiting and to provide some indication as to the
type of wait. Since the reasons for a wait are too numerous to handle individually,
they are grouped into sets of functionally related waits. For each group,
both the number of waits and time the job spent waiting are reported. The
QAPMJOBWTD file provides a description of the type of wait conditions for
each counter set.
Although the file contains fields for up to 16 sets of counters, not all may be used. The number of counters is reported in field JWTNUM. If the number of instrumented counter sets is greater than 16, an additional record or records are written for each reported job.
User
of this file should be aware of the dynamic nature of the content of this
file. Counter sets can be added or redefined by the new release of the operating
system. In addition, IBM® service representatives can define new counter sets
or redefine existing counter sets to allow more granular or more specialized
view of the job wait statistics. As a result, user cannot assume that the
content of this file is always the same.
| Field Name | Description | Attribute |
|---|---|---|
| INTNUM | Interval number: The nth sample database interval based on the start time specified in the Create Performance Data (CRTPFRDTA) command. | PD (5,0) |
| DTETIM | Interval date (yymmdd) and time (hhmmss): The date and time of the sample interval. | C (12) |
| INTSEC | Elapsed interval seconds: The number of seconds since the last sample interval. | PD (7,0) |
| DTECEN | Century digit: where 0 indicates 19XX and 1 indicates 20XX. | C (1) |
| JWNAME | Job / task name. | C (16) |
| JWUSER | Job user. | C (10) |
| JWNBR | Job number. | C (6) |
| JWTDE | System task identifier. | C (8) |
| JWCURT |
Current Wait Time: The time in milliseconds the job has spent in the
current wait. See JWCURB field for the counter set (bucket) which will be
updated when the current wait completes. (This time is not included in other
wait counter sets). |
B (9,0) |
| JWCURE | Reserved. | B (9,0) |
| JWCURB |
Current Counter Set (bucket): If this field is non-zero, it reflects
the counter set that will be updated when the wait completes. |
B (4,0) |
| JWDSEQ | Description sequence number: Identifies the QAPMJOBWTD records associated with this wait data. | B (4,0) |
| JWBGN |
Beginning number of the first counter set reported in this record.
(see Note 1) |
B (4,0) |
| JWCT01 | Count 1. The number of times the job encountered wait conditions associated with this group. | B (9,0) |
| JWTM01 | Time 1. The time in milliseconds the job spent waiting within this group. | B (9,0) |
| JWCTnn JWTMnn |
Count and time are repeated for up to 16 counter sets per record. The
first record reports counter sets 1 to 16. If there are more than 16 counter
sets, the second record for the same job reports counter sets 17 to 32. (also
see Note 1) |
(B (9,0) + B (9,0) ) * 15 |
- When QAPMJOBWT file was created from data collected on a system with operating
system version prior to i5/OS™ V5R4,
up to 32 counter sets can be reported, and all reported counter sets should
be considered to obtain the complete set of wait state data for the job.
When data was collected on a system with operating system version i5/OS V5R4 or later, only the first 16 counter sets should be used for the complete set of job wait state data.
Counter sets with numbers greater than 16 (reported in the second record for the job) represent additional information about wait conditions and will only appear on a system with service activity in progress. Data for these counter sets should normally be ignored and only used under instruction from IBM service representative.
- Wait counters are updated when a job wakes up from a wait; therefore, counters for a job that is currently waiting will not change and the job will not even be reported if it has done no processing in the interval. When the wait is eventually reported, it may be longer than the interval it was reported in.