6. DATA SOURCES › 6.6 I/O Measurements and Device Activity › 6.6.4 Device Activity And Utilization › 6.6.4.2 Device Activity by Device Address
6.6.4.2 Device Activity by Device Address
The CA MICS BATWDA file provides the capability to analyze
the I/O activity of individual devices at the I/S facility.
Each observation in the file represents I/O activity for a
particular device address. The device class, EXCPs, and
device connect time are part of each BATWDA observation. The
file is summarized by performance group (PERFGRP), SYSID, and
HOUR to allow the study of different workloads at different
times.
The BATWDA file accounts for two sets of device classes. One
set consists of the true device classes (like DASD and TAPE).
The other set consists of "pseudo" device classes (like
VIOPAGE and SWAPSSEQ).
The EXCP segments in the raw SMF session, step, and interval
records are the source of the BATWDA observations for the
true device classes. CA MICS summarizes the EXCP segments by
device address as part of the process of building the
PGMAxxxx, PGMNxxxx, PGMMXAxx, and PGMMXNxx program level data
elements discussed in segments 6.6.4.3 and 6.6.4.4. As part
of this process, CA MICS checks the SYSID of the step or
session record. If the WDA option was specified in
prefix.MICS.PARMS(SMFOPS), then observations are produced for
each device address that had I/O activity. Device addresses
that were allocated but had no I/O activity are not included
in the BATWDA file. The following true device classes are
accounted for by this EXCP segment processing:
DEVCLASS Description
-------- -----------
TAPE - Cartridge and Round Reel Tape Device
DASD - DASD Device
COMMGEAR - Communications and Teleprocessing Device
CTC - Channel to Channel Adapter
GRAPHICS - Graphics Device (e.g., a local 3278)
UNIT-REC - Unit Record (e.g., a card reader). This
is kind of a catch-all category containing
many unusual devices, such as the 3838
Vector Processor.
VIO - Virtual I/O (DEVADDR always equals "VIO")
The BATWDA file includes a number of "pseudo" device classes
produced from other segments of the session, step, and
interval SMF records. The amount of activity associated
with these pseudo device classes is recorded in the WDAEXCPS
data element. Unlike the true device class BATWDA
observations, there is no device address to populate the
DEVADDR data element. The DEVADDR data element for these
pseudo device classes contains an abbreviated description of
the pseudo device class. The following pseudo device
classes are accounted for in the BATWDA file.
PSEUDO PSEUDO
DEVCLASS DEVADDR Description
-------- ------- -----------
CMNPAGES CMN - Common Area Page Ins
DEMPAGE DPG - Non VIO, Non-Swap Page Ins +
Non VIO, Non-Swap Page Outs
PAGIN INS - Non VIO, Non-Swap Page Ins
SWAPIN SIN - Pages Swapped In
SWAPPAGE SPG - Pages Swapped In +
Pages Swapped Out
SWAPSSEQ SWP - Address Space Swap Sequences
VIOPAGE VPG - VIO Page Ins + VIO Page Outs
Each DETAIL BATWDA observation contains the start and end
timestamp of the SMF step, session, or interval record that
was its source. The device activity is evenly apportioned
across each HOUR of the duration. Finally, CA MICS
summarizes the BATWDA records to the DAYS timespan. The
BATWDA file only exists temporarily at the DETAIL timespan
during the CA MICS DAILY update because of the enormous
amount of DASD space required.
The method of device activity apportionment over the step,
session, or interval is shown in the following example:
*** EXAMPLE ***
A batch job step runs for 2 hours. The start time is 9:40
and the end time is 11:40.
|--------|--------------------|---------------|
| 16.67% | 50.00% | 33.33% |
| 20 min | 60 min | 40 min |
| HOUR 9 | HOUR 10 | HOUR 11 |
|--------|--------------------|---------------|
| | | |
9:40 10:00 11:00 11:40
CA MICS has no way of knowing exactly when each device
activity occurred, only that it occurred sometime between the
start and end of the step. Therefore, the activity is
apportioned over each hour spanned by the step based on the
percentage of the total step duration. In the example above,
each device address has 16.67% of its activities (ex., EXCPs
Connect Time) assigned to HOUR 9, 50% assigned to HOUR 10,
and 33.33% assigned to HOUR 11.
It should be clear that as step durations become shorter,
the ability to pinpoint the time of device activity
increases. Activation of SMF type 30 interval recording
significantly improves the accuracy of the device activity
data in the BATWDA file.
CA MICS management support applications use the BATWDA file
to give a complete picture of device and "pseudo" device
activity at the I/S facility.
