6. DATA SOURCES › 6.6 I/O Measurements and Device Activity › 6.6.4 Device Activity And Utilization › 6.6.4.4 Data Set Allocations By Device Class
6.6.4.4 Data Set Allocations By Device Class
CA MICS provides a number of data elements estimating data
set allocations in the BATPGM, BAT_ST, and BAT_TS files.
Each observation in these files represents a complete program
step, TSO session, or interval if SMF type 30 interval
recording is active. These data set allocation data elements
are derived by counting the number of SMF EXCP segments by
device class. The complete set of CA MICS data set
allocation data elements in the program-level files are:
PGMA3480 - Cartridge Tape Data Set Allocations
PGMACOMM - Communication Data Set Allocations
PGMADASD - DASD Data Set Allocations
PGMAMDAS - Mountable DASD Data Set Allocations
PGMAGRAF - Graphics Data Set Allocations
PGMATAPE - Magnetic Tape Data Set Allocations
PGMAUREC - Unit Record Data Set Allocations
PGMAVIO - VIO Data Set Allocations
and
PGMMXACA - Max Cartridge Data Set Allocations
PGMMXACO - Max Communication Data Set Allocations
PGMMXADA - Max DASD Data Set Allocations
PGMMXAGR - Max Graphic Data Set Allocations
PGMMXAMD - Max Mountable DASD Data Set Allocations
PGMMXATA - Max Tape Data Set Allocations
PGMMXAUR - Max Unit Record Data Set Allocations
PGMMXAVI - Max VIO Data Set Allocations
These data elements represent data set allocations by device
class. Note that there are two "subclasses" represented:
PGMA3480 and PGMMXACA represent data set allocations to
cartridge tape devices only. PGMATAPE and PGMMXATA represent
data set allocations to all tape devices, including cartridge
types. The same is true for PGMAMDAS and PGMMXAMD. These
data elements represent data set allocations to mountable
DASD only. PGMADASD and PGMMXADA represent all DASD data set
allocations, including mountable DASD.
The data set allocation data elements are grouped into two
very similar looking sets. The difference is that the upper
set, PGMAxxxx, is CA MICS accumulated data elements, while
the lower set, PGMMXAxx, is CA MICS maximum data elements.
At the DETAIL timespan, there is no difference between the
sum and max data elements, unless SMF type 30 interval
recording is active. When interval recording is active, and
a program step, started task, or TSO session generates more
that one interval record, the PGMAxxxx data elements lose
much of their value.
If SMF type 30 interval recording is active, PGMAxxxx
represents actual data set allocations in the first interval
of a step only. In the second and later intervals of a step,
PGMAxxxx is set to zero if the number of data sets is less
than or equal to the highest count of data sets found in a
prior interval. If the number of data sets in the current
interval is greater than the highest prior number
encountered, then PGMAxxxx is set equal to the difference
between the current value and the previous high value. Thus,
the sum of PGMAxxxx over all intervals of a step will equal
the maximum number of CLASS xxxx data sets found in any one
interval of the step.
Why are the values of the PGMAxxxx data elements changed to
zero in the second and later intervals? Because they are
CA MICS accumulated data elements. They are summed to the
job-level JOBAxxxx data elements. The algorithm in place to
modify the true values of data set allocations is there to
protect clients that charge based on the JOBAxxxx data
element values. The switch to SMF type 30 interval recording
would significantly increase the value of JOBAxxxx for a job
with steps generating multiple interval records per step.
This is best explained with an example:
A one-step batch job runs for 2 hours and 23 minutes. The
job step allocates 22 data sets, all residing on 3390 DASD
devices. The data sets remain allocated for the duration of
the step. SMF type 30 interval recording is not active.
BATPGM file Accumulated Elements | Maximum Elements
|
PGMADASD | PGMMXADA
STEP 1 (02:23) 22 | 22
PGMADASD equals PGMMXADA because interval recording is not
active.
Now look at the BATPGM observations from the same job with
SMF type 30 interval recording active and set for 30-minute
intervals.
BATPGM file Accumulated Elements | Maximum Elements
|
PGMADASD | PGMMXADA
INTERVAL 1 (00:30) 22 | 22
INTERVAL 2 (00:30) 0 | 22
INTERVAL 3 (00:30) 0 | 22
INTERVAL 4 (00:30) 0 | 22
INTERVAL 5 (00:23) 0 | 22
With interval recording active, the value of PGMADASD has
been set to zero in the second through the fifth intervals.
This was done to avoid a dramatic increase in the job-level
data element JOBADASD. Without the algorithm that modifies
the PGMAxxxx data elements in the second and greater
intervals, JOBADASD would have changed from 22 to 110 with no
change in the executing job step.
If SMF type 30 interval recording is active, the PGMMXAxx
data elements should be used for data set allocation analysis
at the DETAIL timespan. Without interval recording, the
maximum and accumulated data elements are identical at the
DETAIL timespan.
The following limitations are associated with the PGMMXAxx
and PGMAxxxx data elements:
o These data elements estimate data set allocations by
summing the number of EXCP segments by device class.
The fact that a data set was allocated does not imply
nor guarantee that it was used.
o There are a number of factors that will cause the
PGMAxxxx and PGMMXAxx values to overstate the number of
data sets allocated by device class.
If SMF type 30 interval accounting is active, interval
records are produced for every step or session, even if
the step or session executes for less than the specified
interval time. If the interval is set for 30 minutes
and a job step runs for 2 minutes, an SMF type 30
subtype 3 ("partial" or "last" interval) record is
produced for the step. SMF will also produce an SMF
type 30 subtype 4 (step end) record for the step. In
theory, these two records should have identical EXCP
segments. A difference in the method used by SMF to
consolidate the EXCP segments in interval versus the
step-end records, however, allows up to two EXCP
segments with the identical DDNAME and device address to
appear in the interval records while the step-end record
will only have a single segment. This difference only
occurs when data sets are dynamically allocated,
unallocated, and reallocated to the same DDNAME and
device address within the interval. This lack of
agreement in the number of EXCP segments does not affect
the numeric quantities in the records, such as EXCPs and
device connect time. It only affects the PGMAxxxx and
PGMMXAxx data elements that depend on the number of EXCP
segments present. CA MICS by default uses the subtype 3
rather than the subtype 4 record when both are
available. This means that the values of the PGMAxxxx
and PGMMXAxx data elements will increase when interval
recording is turned on, even for steps and sessions that
execute in less than the specified interval time.
Multi-volume data sets are another factor that
exaggerates the value of these data elements. When a
data set resides on several DASD volumes, each volume
will have its own EXCP segment in the SMF record. A
single tape data set, spanning multiple tape volumes,
will also generate multiple EXCP segments if the tape
volumes are mounted on different devices.
If the same data set is allocated by more than one
DDNAME in the JCL, each DDNAME will have its own EXCP
segment in the SMF record.
Finally, if the DDCONS(NO) keyword is specified in the
SMF PARMLIB member, EXCP segment consolidation will not
be performed by SMF. This can dramatically increase the
number of EXCP segments produced for all SMF type 30
records for long running steps that perform thousands of
dynamic allocations and unallocations. A corresponding
increase in the PGMAxxxx and PGMMXAxx values will occur.
o A data set that is dynamically allocated or deallocated
during an interval will count as a PGMMXAxx value for
that interval. A deallocated device will be noted as
absent in the subsequent interval record.
The ability of CA MICS to determine when data sets were
allocated by a job step is hampered because there is no
indication of wall clock connect time in the EXCP segments.
The activation of SMF type 30 interval records can minimize
this uncertainty.
The data elements representing device allocations in the
BATJOB and BAT_JS files are derived from the corresponding
program-level elements. These data elements are:
JOBA3480 - Cartridge Tape Data Set Allocations
JOBACOMM - Communication Data Set Allocations
JOBADASD - DASD Data Set Allocations
JOBAGRAF - Graphics Data Set Allocations
JOBAMDAS - Mountable DASD Data Set Allocations
JOBATAPE - Magnetic Tape Data Set Allocations
JOBAUREC - Unit Record Data Set Allocations
JOBAVIO - VIO Data Set Allocations
and
JOBMXACA - Max Job Cartridge Data Set Allocations
JOBMXACO - Max Job Communication Data Set Allocations
JOBMXADA - Max Job DASD Data Set Allocations
JOBMXAGR - Max Job Graphic Data Set Allocations
JOBMXAMD - Max Job Mountable DASD Data Set Allocations
JOBMXATA - Max Job Tape Data Set Allocations
JOBMXAUR - Max Job Unit Record Data Set Allocations
JOBMXAVI - Max Job VIO Data Set Allocations
These job-level data elements are useful for determining jobs
that allocate large numbers of data sets. The JOBAxxxx data
elements are CA MICS accumulated elements. At the DETAIL
timespan, they represent the sum of the corresponding
PGMAxxxx data elements from all steps of the job. At higher
timespans, they represent the sum of all occurrences of
JOBAxxxx that were included in the level of summarization.
The JOBMXAxx data elements are CA MICS maximum data elements.
At the DETAIL timespan, they represent the maximum value of
the corresponding PGMMXAxx data elements encountered in all
steps or intervals of the job.
The differences in these accumulated and maximum data
elements are best described with an example:
A three-step batch job runs for 25 minutes.
o Step one allocates 46 DASD data sets for the duration of
the 10-minute step.
o Step two allocates 26 DASD data sets for the duration of
the 5-minute step.
o Step three allocates 55 DASD data sets for the duration of
the 10-minute step.
|---------------------|----------|--------------------|
| 46 DASD data sets | 26 DASD | 55 DASD data sets |
| | data sets| |
|---------------------|----------|--------------------|
|<---- STEP 1 ------>|<-STEP 2->|<---- STEP 3 ----->|
BATPGM file Accumulated Elements | Maximum Elements
|
PGMADASD | PGMMXADA
Step 1 (00:10) 46 | 46
Step 2 (00:05) 26 | 26
Step 3 (00:10) 55 | 55
BATJOB file |
JOBADASD | JOBMXADA
Job (00:25) 127 | 55
The value of JOBADASD is 127. It is known that approximately
127 DASD data sets were allocated sometime over the three
steps of the job.
The value of JOBMXADA is 55. It is known that a maximum of
55 DASD data sets were allocated in any one interval or step
of the job.
Copyright © 2014 CA.
All rights reserved.
|
|