SirAud report contents
Overview
Each line in a PERFREPT report corresponds to a partial interval as defined by the ACCTIM parameter. The columns in the report are derived from the information in the system partial statistics entries, since last statistics entries for CMPL, LOAD, and EVAL activities, and user logout statistics entries. Unless otherwise noted, statistics values come from the system partial statistics entry.
| Column | Content |
|---|---|
| DATE | The date when the statistics for the interval were logged to the journal. |
| END OF INTVL | The time when the statistics for the interval were logged to the journal. |
| AVG USERS | The connect time (CNCT) used by all logged in users during the interval, divided by the seconds in the interval. The connect time for a user is reported in the user partial and logout statistics entries. |
| %CPU | The CPU value divided by the CNCT value. |
| PCPU | The PCPU value reported for the interval
when Model 204 is running on a single CPU. When MP is active and offload subtask partial statistics are produced, the PCPU for the interval is the CPU for the interval divided by the PR for the interval. |
| SCREENS/SEC, DKRD+DKWR/SEC, SVRD+SVWR/SEC, LKWAIT+LKPOST/SEC | The statistics or sum of statistics, divided by the number of seconds in the interval. |
| JRNL I/O/SEC | The number of blocks written during the interval, divided by the number of seconds in the interval. |
| RESP TIME SECS/SCRN MEAN | The mean response time in seconds per screen over the interval.
Sample points for determining the mean are gathered from the user since last statistics entries. Each execution of a procedure is accompanied by a CMPL/EVAL or LOAD/EVAL since last pair. The RQTM statistic reported in the since last entries is summed for a user until an EVAL since last entry is found that has a SCREENS value not equal 0. A sample point is then determined by taking the RQTM sum and dividing it by SCREENS. In the case when a partial interval completes and a sample point is in progress (no EVAL with SCREENS not equal 0 found yet), the sample point is determined using the RQTM sum through the last EVAL and dividing it by 1. The RQTM from the last CMPL or LOAD since last entry is attributed to the next partial interval. |
| RESP TIME SECS/SCRN STDEV | The standard deviation of the response time in seconds per screen over the interval. Sample points for determining the standard deviation are the same as with the mean. |
| CPU SECS/SCRN MEAN | The mean CPU time used in seconds per screen. Sample points for determining the mean are calculated the same way as for the mean response time, except that the CPU statistic is used instead of RQTM. |
| CPU SECS/SCRN STDEV | The standard deviation of the CPU time used in seconds per screen over the interval. Sample points for determining the standard deviation are the same as with the mean. |
| DKRD+DKWR/SCRN, SVRD+SVWR/SCRN | Those statistics values divided by the SCREENS for the interval. |
FILEREPT
This report is a subset of statistics from the file close statistics entries and other file information reported in the journal, totalled for each file and for all files. The Model 204 files are listed in descending order by the sum of the DKRD and DKWR columns.
| Column | Content |
|---|---|
| FILE NAME | The name of the Model 204 file. The first line of the report presents the totals for all Model 204 files and uses *TOTALS* as the name. |
| DKRD, DKWR, DIRRCD, BXFIND, BXNEXT, BXRFND, BADD+BDEL+BCHG, IXADD+IXDEL, RECADD+RECDEL, BXINSE+BXDELE+BXCHNG, RETRYC | The totals of the specified statistic or sum of statistics. |
| DKUPDT | The number of times pages from the file were flushed out of the buffer pool. |
PROCREPT
One logical line is produced for each unique combination of procedure name, procedure file name, and subsystem name executed during the Model 204 run, and a summation line is produced for all procedures executed. Each logical line requires two physical lines.
The columns are derived from user since last statistics entries for CMPL,
LOAD, or EVAL activities.
The procedures are sorted in descending order by TOTAL CPU SECONDS.
| Column | Content |
|---|---|
| RANK | The rank of the procedure by TOTAL CPU SECONDS. |
| PROCEDURE NAME | The name of the procedure. |
| PROC FILE | The name of the file to which the procedure belongs. |
| SUBSYSTEM NAME | The name of the subsystem to which the procedure file belongs. |
| NUMBER OF RUNS | The number of times a procedure was executed. |
| TOTAL CPU SECS | The total CPU from the CMPL, LOAD, and EVAL since last entries converted from milliseconds to seconds. |
| EVAL CPU SECS | The total CPU from EVAL since last entries converted from milliseconds to seconds. |
| STCPU/CPU | The sum of all STCPU divided by the sum of all CPU. |
| RESP TIME MEAN | The mean response time in seconds per procedure run. An execution of a procedure is accompanied by a CMPL/EVAL or LOAD/EVAL since last pair, and it determines a sample point for calculating the mean response time.
SirAud assigns a variable known as RUNIN to each run. RUNIN is set to the SCREENS from the EVAL since last entry, if SCREENS is not equal to 0. Otherwise, RUNIN is set to the IN from the EVAL since-last entry, if IN is not equal 0. Otherwise, RUNIN is set to 1. The RQTM statistic from the CMPL or LOAD since last entry is added to the RQTM from the EVAL. A sample point is then determined by taking the RQTM sum and dividing it by RUNIN. |
| RESP TIME STDEV | The standard deviation of the response time in seconds per procedure run. Sample points for determining the standard deviation are the same as with the mean. |
| TOTAL DKRD+DKWR, TOTAL BXFIND, TOTAL BXNEXT |
The totals of the specified statistic or sum of statistics. |
| UPDATE UNIT SECS MEAN | The mean time in seconds an update unit was active for the procedure. A sample point is the time between the start and end of an update unit. |
| UPDATE UNIT SECS STDEV | The standard deviation of the number of seconds seconds an update unit was active for the procedure. Sample points for determining the standard deviation are the same as with the mean. |
| NUMBER OF SCREENS | The total SCREENS for the totals line. The sum of the RUNIN variable for a procedure is used for an individual procedure. |
| AVERAGE CPU SECS | The total CPU divided by NUMBER OF SCREENS, then converted from milliseconds to seconds. All other "AVERAGE" columns are the total of the specified statistic divided by NUMBER OF SCREENS. |
| AVERAGE DKRD+DKWR, AVERAGE DKPR, AVERAGE SVRD+SVWR, AVERAGE STDEQ, AVERAGE FINDS, AVERAGE DIRRCD, AVERAGE RECDS, AVERAGE SORTS, AVERAGE STRECDS | The totals of the specified statistic or sum of statistics divided by NUMBER OF SCREENS. |
SUBREPT
There is one line produced for each subsystem that was active during the Model 204 run and a summation line for all subsystems that were active. The columns are derived from user since last statistics entries for CMPL, LOAD, or EVAL activities for member procedures. The subsystems are sorted in descending order by TOTAL CPU SECONDS.
| Column | Content |
|---|---|
| SUBSYSTEM NAME | The name of the subsystem. The first line of the report presents the totals for all subsystems and uses *TOTALS* as the name. *NONE* is used for reporting on procedures that were run outside of subsystem context. |
| TOTAL CPU SECS | The total CPU from the CMPL, LOAD, and EVAL since last entries converted from milliseconds to seconds. |
| EVAL CPU SECS | The total CPU from EVAL since last entries converted from milliseconds to seconds. |
| STCPU/CPU | The sum of all STCPU divided by the sum of all CPU. |
| TOTAL DKRD+DKWR, TOTAL SVRD+SVWR | The total of the specified statistics sums. |
| NUMBER OF SCREENS | The total SCREENS from EVAL since last entries. |
| NUMBER OF RUNS | The number of times member procedures were executed. |
| RESP TIME MEAN | The mean response time in seconds per member procedure run. An execution of a member procedure is accompanied by a CMPL/EVAL or LOAD/EVAL since last pair, and it determines a sample point for calculating the mean response time.
SirAud assigns a variable known as RUNIN to each run. RUNIN is set to the SCREENS from the EVAL since last entry, if SCREENS is not equal 0. Otherwise, RUNIN is set to the IN from the EVAL since last entry, if IN is not equal 0. Otherwise, RUNIN is set to 1. The RQTM statistic from the CMPL or LOAD since last entry is added to the RQTM from the EVAL. A sample point is then determined by taking the RQTM sum and dividing it by RUNIN. |
| RESP TIME STDEV | The standard deviation of the response time in seconds per member procedure run. Sample points for determining the standard deviation are the same as with the mean. |
| CPU SECS/SCRN | CPU converted to seconds from milliseconds and divided by SCREENS. All other "/SCRN" columns are the statistic value divided by SCREENS. |
| DKRD+DKWR/SCRN, SVRD+SVWR/SCRN | Sums of statistics divided by the SCREENS. |
| SERVER SIZE (K) | The smallest server size, in units of 1024 bytes, needed to execute the member procedures without any UTABLE commands. A server is made up of a fixed portion, determined by Model 204 user initialization parameters (for example, LOBUFF) that are fixed during the run, and a variable portion containing tables that can be resized using a UTABLE command. The size calculated for this column is based on the member procedures that were actually executed during the run. See UTBLREPT for a more detailed explanation. |
UTBLREPT
This report provides a guide to how large a server is actually needed to run a site's application subsystems. There is one set of lines produced for each subsystem that was active during the Model 204 run. There is a column for each server table that can be reset with a UTABLE command. The values are derived from since last statistics entries for CMPL, LOAD, or EVAL activities for each procedure that is a member of the subsystem. The value reported is the highwater mark over all the entries.
SirAud calculates one fixed server size using table sizes specified on the Model 204 user zero parameter line, the IODEV parameter lines, or the Model 204 system default. The largest value encountered is used in the calculation. This is different from the way Model 204 operates in that SirAud uses one fixed server size for reporting purposes while Model 204 calculates a fixed server size for each user. The variable server size is calculated for each subsystem using the aforementioned server table highwater marks from the since last statistics entries.
| Column | Content |
|---|---|
| SERVER SIZE (K) | The sum of the fixed and variable server sizes in units of 1024 bytes. This value shows approximately the largest server needed to execute all the subsystem's procedures that were executed during the Model 204 run.
The subsystems are sorted in descending order by SERVER SIZE (K) for the line identified by The columns other than SERVER SIZE (K) are the highwater marks of the corresponding server table as reported in the since last entries for the member procedures that were executed. These columns are: LFTBL, LGTBL, LITBL, LNTBL, LQTBL, LSTBL, LTTBL, LVTBL, LXTBL, LFSCB, and LPDLST. |
| SUBSYSTEM | The name of the subsystem. The first line of the report presents the totals for all subsystems and uses
*TOTALS* as the name. *NONE* is used for reporting on procedures that were run outside of subsystem context. |
| PROCEDURE NAME | The name of the member procedure. |
| PROC FILE | The name of the file to which the member procedure belongs. |
The SirAud parameter UTBLRANG is used to identify member
procedures that possibly cause the subsystem to require a larger server.
For each server table, SirAud calculates the mean highwater
and standard deviation around the mean over all the member
procedures that were executed.
UTBLRANG specifies the number of standard deviations an individual procedure server table highwater
mark can stray from the subsystem mean before it is flagged out of range.
Once all server tables of all member procedures have been
examined, SirAud computes a new set of highwater marks and server size
using just those procedures whose server table highwater marks are in range.
These values are reported on the line identified by *TARGET*.
The "problem procedures" are then reported individually, sorted by the server size required to execute the
procedure and its server table highwater marks.
UTBLSUMM
This report presents the server table highwater
marks that are used to produce the *CURRENT* lines of UTBLREPT.