Interpret the Process Statistics

Evaluate Performance in Detail › Performance Evaluation Report › Interpret the Process Statistics

Interpret the Process Statistics

The following example illustrates the default process statistics section of the Performance Evaluation Report.

The /PROCESS_STATISTICS qualifier allows you to tailor the process statistics section of the Performance Evaluation Report. You can specify the focus of the report to obtain different sets of statistics that pertain to the focus area. The grouping, merging, and sorting of the process data is controlled with the primary and secondary key settings.

The following list shows the primary and secondary keys:

MODE
USERNAME
IMAGENAME
UIC_GROUP
PROCESS_NAME
WORKLOAD_NAME
ACCOUNT_NAME
PID

For more information on how to specify the /PROCESS_STATISTICS qualifier, see the chapter Performance Manager Commands.

To display only the process statistics section of the Performance Evaluation Report, use the following qualifier:

/INCLUDE=PROCESS_STATISTICS

To disable the process statistics display from the Performance Evaluation Report, use the following qualifier:

/INCLUDE=NOPROCESS_STATISTICS

Because process classification by PID or PROCESSNAME results in virtual memory requirements, these reports keys are disabled by default and require you to specifically enable them. For more information on virtual memory requirements, see the appendix Estimate Virtual Memory Needs.

The following example shows a Performance Evaluation Report, Process Statistics by Image for Interactive, Batch, Detached, and Network Jobs:

Performance YQUEM (VAX 6000-440) Page 1 Evaluation PA Vx.x Tuesday 26-JAN-1997 09:00 to 10:00 +--------------------------------------------------------------------+ | The table below lists observed workload characteristics of all the | | interactive images that were run during the given interval. Note | | that Diskio, Bufio and Cputim are percentage contributions of the | | respective images to the total system load. Working set size and | | working set faults are the average for the respective images. In | | the case of 0 image activations, the Uptime/image and Cputim/image | | actually report the cumulative Uptime and Cputim for the image. | +--------------------------------------------------------------------+ Node Name: YQUEM MODE: INTERACTIVE # of Page Faults Avg. % of % of Uptime/ Cputim/ activ- per Actvtn Ws Direct Buffered % of image image Image ations -Soft--Hard size I/O I/O Cputim (sec) (sec) -------- ------- ------ ---- ------ ------ ------ ------ ------- -------- 1 2 3 4 5 6 7 8 9 10 (dcl) 0 830 0 492 0.05 0.23 0.11 142470 16.54 ACS 1 242 13 818 0.00 0.01 0.00 2 0.29 CDU 1 8531 5 2168 0.01 0.01 0.01 3 2.00 CLR 1 68 7 345 0.00 0.01 0.00 6 0.36 CMS 5 522 31 509 0.12 0.13 0.06 48 1.59 COPY 5 124 3 540 0.33 0.15 0.02 16 0.43 DEBUGSHR 2 1739 44 1938 0.01 0.06 0.03 724 2.10 DECPRESENT 2 17125 138 5185 2.34 4.04 1.95 1774 140.73 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VMOUNT 1 107 13 520 0.00 0.18 0.00 361 0.62 VMSHELP 4 132 7 660 0.04 0.11 0.01 31 0.40 VTX$CLIENT_C 3 540 20 916 0.13 0.17 0.03 98 1.62 ---- ------- ------- ------ Totals 197 24.70 29.94 12.30 11 12 13 14

Performance YQUEM (VAX 6000-440) Page 5 Evaluation PA Vx.x Tuesday 26-JAN-1997 09:00 to 10:00 +------------------------------------------------------------------------+ | The following table summarizes the workload characteristics on a per | | image activation basis. Note that values would be zeros if total | | number of image activations is zero. | +------------------------------------------------------------------------+ PrimaryKey: Avg. Avg. Avg. Avg. Mode # of Avg. Soft Hard Direct Buff'd Avg. Images Secondary Key: processes WSiz/ flts/ flts/ IO/ IO/ Cputim/ per None activ/inact image image image image image image Second -------------- ----- ----- ----- ----- ----- ------ ------ ------- -------- 15 16 17 18 19 20 21 22 23 24 INTERACTIVE 32 53 1905 830.8 15.7 122.8 473.3 8.99 0.0547 BATCH 1 1 4144 0.0 0.0 0.0 0.0 0.33 0.0000 NETWORK 5 2 671 367.4 4.6 35.0 114.9 1.19 0.0389 DETACHED 31 38 1855 1592.0 9.0 5295.3 15542 124.73 0.0036

The following statements are keyed to the columns in the previous example:

By default, the Performance Manager displays the process information by image name. In this example, images running or waiting on the system during the report time period are shown.
Number of times that an image was activated during the report time period. If an image has zero activations, than it has been activated previously (before the reporting period). If you specify the secondary key as USERNAMES, the Performance Manager displays the number of image activations per user.
Number of soft page faults incurred by an image during the report time period, divided by the number of activations. If you specify the secondary key as USERNAMES, this column displays the total number of soft page faults for all images, divided by the total number of image activations for the user.
Number of hard page faults incurred by an image during the report time period, divided by the number of activations. If you specify the secondary key as USERNAMES, this column displays the total number of hard page faults for all images, divided by the total number of image activations for the user, invoked per user.
Average number of process private pages plus the global pages for this image (or user) during the report time period.
Percentage of all direct I/O attributable to an image or user during the report time period.
Percentage of all buffered I/O attributable to an image or user during the report time period.
Percentage of all CPU time attributable to an image or user during the report time period.
Total elapsed time (wall clock, in seconds) of an image or user, divided by the number of its activations. If the number of image activations is zero, this measurement represents the total residence time of all activations of the image (or all images if the USERNAMES option was specified).
Total amount of CPU seconds used by processes running an image (or by a user if the USERNAMES option was specified) during the report time period, divided by the number of its activations, unless the number of activations is zero.
Total number of image activations due to interactive, batch, or network processes, calculated for the report time period. This example shows 197 interactive image activations.
Percentage of all direct I/O due to interactive, batch, or network processes during the report time period. In this example, 24.70 percent of all direct I/O was due to interactive processes.
Percentage of all buffered I/O due to interactive, batch, or network processes during the report time period. In this example, 29.94 percent of all buffered I/O was due to interactive processes.
Percentage of all CPU time used by interactive, batch, or network processes during the report time period. In this example, interactive processes consumed 12.30 percent of all CPU time.
Process type: interactive, batch, network or detached, or name of workload when primary key options are used or /CLASSIFY_BY.
Average number of active processes. In this example, there is an average of 32 active interactive processes during the 30 intervals.
Average number of inactive processes. In this example, there is an average of 53 inactive interactive processes during the 30 intervals.
Average number of private and global pages in the process's working set for the active processes.
Average number of soft page faults calculated by dividing the total number of soft page faults (for this type) by the number of image activations. A soft page fault is the total number of times that processes reference a virtual page that is not in its working set but is in memory.
Average number of hard page faults calculated by dividing the total number of hard page faults (for this type) by the number of image activations. A hard page fault is the total number of times that processes reference a virtual page that is not in its working set and requires a read operation from disk.
Average number of direct I/O operations per image. Calculated by dividing the total number of direct I/O operations (for this type of process) by the total number of image activations.
Average number of buffered I/O operations per image. Calculated by dividing the total number of buffered I/O operations by the total number of image activations. Buffered I/O operations use intermediate system buffers rather than process context buffers.
Average CPU time used per image. Calculated by dividing the total CPU time accrued by processes, in seconds, by the total number of image activations.
Images per second. Total number of image activations divided by the total elapsed wall-clock time during which processes were active, resulting in the average number of images completed per second.