Viewing the result of a long-term analysis

Follow the procedure to view the results of a long-term analysis.

About this task

When the long-term analysis wizard finishes, the Buffer Pool Analyzer main window shows the result in one of the Results subfolders of the Long-Term Analysis folder. The subfolders can contain results from several long-term analyses. The result from the most recent analysis is highlighted.
Figure 1. Long-Term Analysis – The Results Selection window
This figure is a graphical representation of the Results Selection window from a long-term analysis.

Results are named <analysis_type>-<subsystem>-<date> <time>, whereby <analysis_type> and <subsystem> correspond to your specifications in Step 2: Choosing a subsystem and specifying an analysis type, and <date> and <time> stand for the date and time when the long-term analysis result was generated and saved.

Procedure

  1. If you want to delete results from the folder, select a specific result by clicking it. Then press the Delete key. To delete all results, right-click Results. Then click Delete all. You are asked to confirm the deletion.
    Note: Results remain on the hard disk drive and take up space until they are deleted. They are usually located in folder C:\Documents and Settings \<userid> \db2pev<version> \bpa-zos-reports \longterm-analysis. However, because of their special format, do not manipulate the folder contents manually.
  2. To view the result from a long-term analysis, double-click it, or select it and press Enter.

    The result is shown in the right pane of the Buffer Pool Analysis main window. The result consists of a chart and a corresponding legend and report. The legend contains symbols and text that explain the chart. The report lists the information in table form and shows the values represented in the chart. The legend and the report can be switched on or off by using the Legend and Report push buttons.

    All results can also be shown in your web browser. Right-click into a graphic and choose Open in browser.

    The long-term analysis function generates results that differ depending on the analysis type that is specified in Step 2: Choosing a subsystem and specifying an analysis type. The following list shows and describes examples of charts from each analysis type to help you understand how your specifications (mainly the counters and objects) and the performance data from the bpd files are reflected in the result.

What to do next

A Weekly view by day analysis result:
Figure 2. Long-Term Analysis – Example of Weekly view by day result
This figure is a graphical representation of an example of a Weekly view by day result from a long-term analysis.

This analysis type shows counter values per weekday of selected counters and objects. Counter values represent per-minute values, for example 5 000 Getpage total operations per minute on average over a day. One counter value per counter, object, and weekday is shown, for example, 3 000 Read page operations (the counter) on BP0 (the object) on Monday (the weekday), 2 500 for the same combination on Tuesday, and so on. The counter values for the seven weekdays are connected by lines for better readability. (The lines themselves do not represent interim values.) The Average counts show the calculated averages over all affected objects for each counter per weekday, for example, the average of the Write page operations (the counter) of buffer pool BP0 and BP1 (the objects). The interpretation of these average counts is only reasonable if the affected objects are of the same type, for example only buffer pools or only page sets. If you selected objects of different types, the average values are calculated over all objects and do not yield helpful results. Note that counters and ratios are treated equally in this graphic, except that they have their dedicated y-axis.

If the data from the bpd file spans several weeks, the values are overlaid, which means that the described graphic for one week is overlaid with a similar graphic for the second week (having different values), and so on. This example clarifies what is already described in Step 3: Specifying counters, objects, time frame, and output: You can easily overload the graphic by selecting too many counters and objects for longer periods.

You can use this analysis type to analyze how certain counters develop over a week (if the time frame covers a week), or to compare how counters develop over several weeks. This type helps to identify counters that show conspicuously high or low values at specific weekdays or show a trend toward lower or higher values over weeks.

A Daily view by hour analysis result:
Figure 3. Long-Term Analysis – Example of Daily view by hour analysis result
This figure is a graphical representation of an example of a Daily view by hour result from a long-term analysis.

This analysis type shows counter values per hour of selected counters and objects. Counter values represent per-minute values, for example 5 000 Getpage operations per minute on average over an hour. One counter value per counter, object, and hour of the day is shown, for example, 7 000 Read page operations (the counter) on BP0 (the object) between 4:00 p.m. and 5:00 p.m. (the hour), 3 000 for the same combination during the next hour, and so on. The counter values for the 24 hours of a day are connected by lines for better readability. (The lines themselves do not represent interim values.) The Average counts show the calculated averages over all affected objects for each counter per hour, for example, the average of the Write page operations (the counter) of buffer pool BP0 and BP1 (the objects). The interpretation of these average counts is only reasonable if the affected objects are of the same type, for example only buffer pools or only page sets. If you selected objects of different types, the average values are calculated over all objects and do not yield helpful results. Note that counters and ratios are treated equally in this graphic, except that they have their dedicated y-axis.

If the data from the bpd file spans several days, the values are overlaid, which means that the described graphic for one day is overlaid with a similar graphic for the second day (having different values), and so on. The same precautions should be taken as with the Weekly view by day analysis to avoid overloaded results.

You can use this analysis type to analyze how certain counters develop over a day (if the time frame covers a day), or to compare how counters develop over several days. It is basically a more detailed analysis than the Weekly view by day analysis.

A View of a period of time analysis result:
Figure 4. Long-Term Analysis – Example of View of a period of time analysis result
This figure is a graphical representation of an example of a View of a period of time result from a long-term analysis.

This analysis type shows counter values of selected counters and objects from several bpd files in chronological order. Counter values represent per-minute values, as in the previous analysis types. One counter value per counter, object, and bpd file is shown, for example, 1 000 Read request operations (the counter) on BP0 (the object) on average from data from the first bpd file, 1 050 for the same counter and object from the second bpd file, and so on. The counter values are connected through lines for better readability. The y-axis on the left side is applicable to counter values, the one on the right side to ratios, if those were selected.

This analysis type provides meaningful information when several bpd files were selected, and if the effect of this selection was not canceled by a restrictive specification of a time frame. For example, if you have selected seven bpd files, whereby each file contains performance data of one subsequent day, but you have restricted the time frame to the second and third day, only these two bpd files are effectively used in this analysis.

Each effectively used bpd file is identified on the x-axis by a timestamp, and the files are shown in ascending order from left to right. The identifying timestamp is the timestamp of the latest performance record found in a bpd file (which might not necessarily be used in this analysis if the time frame restricts the use to some time before the latest record).

A further clarification on the y-values is appropriate: The calculated per-minute values of counters and ratios of effectively used bpd files are based on the specified (or default) time frame. For example, if the latest bpd file contains performance records of one day between 8:00 a.m. and 12:00 a.m., but you have specified a time frame limit of 9:00 a.m. (for whatever reason), the values for selected counters and ratios are calculated based on performance records between 8:00 a.m. and 9:00 a.m. (Nevertheless, the identifying timestamp for this bpd file on the x-axis shows 12:00 a.m.)

You can use this analysis type to analyze how certain counters develop over long periods, by using your portfolio of historical bpd files.

A Bar chart analysis result:
Figure 5. Long-Term Analysis – Example of Bar chart analysis result
This figure is a graphical representation of an example of a Bar chart result from a long-term analysis.

This analysis type shows the distribution of counter values of selected counters over selected objects as bar chart. As usual, counter values are per-minute values. The selected counters in this example are the Getpage total counter, the Read page counter, and the Read request counter. The selected objects are either one or more objects of one or more buffer pools, or all objects of one or more buffer pools, dependent on your selections in Step 3: Specifying counters, objects, time frame, and output. In this example, buffer pools BP0, BP1, and BP2 were selected, which means that the three counters encompass the activities of all objects in these buffer pools. The x-axis reflects the selected objects, here the buffer pools.

You can use this analysis type to easily compare selected counters in selected objects, for example, to compare the workload in selected buffer pools.

A Pie chart: display 1 counter and n objects analysis result:
Figure 6. Long-Term Analysis – Pie chart: display 1 counter and n objects analysis result
This figure is a graphical representation of an example of a Pie chart: display 1 counter and n objects result from a long-term analysis.

This analysis type shows a 1-to-n relationship of a selected counter to several selected objects as pie chart. Each slice of the pie represents one of the selected objects; the size of the slice corresponds to the percentage of the total of all selected objects. In this example, the selected counter is the Getpage total counter, and the selected objects are the objects in buffer pools BP0, BP1, and BP2. The percentages are shown in the graphic; the corresponding values per object (as per-minute values) are shown in the report following the graphic.

You can use this analysis type to compare a few values to a total, for example, to determine how much of the Getpage total activity happens in the most important buffer pools.

A Pie chart: display n counters and 1 object analysis result:
Figure 7. Long-Term Analysis – Pie chart: display n counters and 1 object analysis result
This figure is a graphical representation of an example of a Pie chart: display n counters and 1 object result from a long-term analysis.

This analysis type shows a n-to-1 relationship of several selected counters to a selected object as pie chart. Each slice of the pie represents one of the selected counters; the size of the slice corresponds to the percentage of the total of all selected counters. In this example, the selected counters are Getpage total, Read page, and Read request, and the selected object is BP2. The percentages are shown in the graphic; the corresponding values per counter (as per-minute values) are shown in the report following the graphic.

You can use this analysis type to compare a few values to a total, for example, to determine which counters have the most activity in a buffer pool.