A new way to analyze historical performance data on IBM i
Using Performance Data Investigator Graph History function
Graph History in IBM Navigator for i
Collection Services and Performance Data Investigator have previously allowed you to look at a day's worth of performance data in graphical views. System Monitoring and Performance Data Investigator allow you to monitor live performance data graphically. Now in IBM i 7.3 release there is historical data support with Collection Services and graphical display of that data with Graph History in the Performance task of IBM Navigator for i.
Historical data can help you identify trends and investigate performance issues from the past or to compare today's performance with a time period in the past that you select.
Collection Services keeps up to 24 hours of data in each collection. By default, the collections are preserved on a system for 10 days. To look at performance from week to week, month to month, or even year to year, it was necessary to provide a way to keep performance data for longer periods, without taking up too much storage and also to allow an easy way to view data across multiple days.
Collection Services now does this through the new Historical Data collection. The new performance collection type *HSTFILE contains historical data in a file-based performance collection.
The Historical Data collection is data collected by Collection Services that has been reduced and summarized to keep the data over a long period of time and conserve space. The data is intended to be just a subset of the original Collection Services data.
The Historical Data collection will be populated with data from management collection objects in the configured collection library (for example, QPFRDATA) when the Collection Services collector is cycled or ended.
To enable the concept of aging or decaying data, there are two levels of historical data – detail and summary. As you can see from the following table, as the maximum collection duration increases, the number of metrics available decreases.
Table 1. Collection data duration and variety
|Data||Maximum duration of collection||Number of metrics available|
(variety of data)
|Collection Services data in *CSFILE||24 hours||most|
|Historical Detail data in *HSTFILE||60 days||fewer|
|Historical Summary data in *HSTFILE||50 years||least|
The summary historical data is the system-level and summarized metrics. This data is useful in identifying trends or detecting changes in a system over a long period of time.
The detail historical data is the collected data saved for top contributors of each interval, as well as other relevant supplementary data. This is the supporting data for the summarized metrics. The detail data can be used when looking deeper into a problem identified from summary historical data. Only the top contributors for each metric will be stored as historical detail data.
Create historical data on IBM i 7.3
To turn on the creation of historical data on an IBM i 7.3 partition, use the Collection Services configuration panel. Under the Performance task on IBM Navigator for i, click Configure Collection Services.
Figure 1. Configure Collection Services
On the General tab, select the Create historical data when collection is cycled check box. This check box is cleared by default.
Figure 2. Option to create historical data
To set the historical data retention options, click the Data Retention tab.
Figure 3. Historical data retention settings
On this panel, you can specify the length of time that historical data will be kept. The time span that historical data can be saved depends on whether or not PM Agent is turned on. If PM Agent if set to off, Collection Services will only preserve 7 days of detail historical data and 1 month of summary historical data. With PM Agent set to on, Collection Services will preserve up to 60 days of detail historical data and 50 years of summary historical data. The following table shows the maximum and default settings.
Table 2. Historical Data retention
|Historical Data Retention||Maximum||Default|
|Summary Data with PM Agent off||1 month||1 month|
|Detail Data with PM Agent off||7 days||7 days|
|Summary Data with PM Agent on||50 years||1 year|
|Detail Data with PM Agent on||60 days||7 days|
For more historical data configuration options, click the Historical Data tab.
Figure 4. Historical Data settings
On this panel you can specify the historical data interval (default is 60 minutes). The summary and detail historical data will be saved in the Historical Data collection at this requested interval.
You can use any the following options to configure historical detail data.
- Create historical detail data
You may create only summary historical data or both summary and detail historical data. The default is to create historical detail data when summary historical data creation has been selected.
- Detail data filter
The detail data filter determines what number of top contributors for each metric will be stored in the detail data files. The top contributors for each metric means the job or task, disk, memory pool, or communication line that had the most significant values (CPU %, response time, utilization, and so on) during that interval. Only data for these top contributing elements are saved as historical detail data for that interval. The default is that 20 top contributors for each metric at each interval will be stored.
After you have set the options, click OK at the bottom of the panel. If you have turned PM Agent on when it was previously off, you will need to read and agree to the disclaimer displayed regarding the use of PM Agent.
Note: Turning PM Agent on does not cause data to be transmitted to IBM. The user must still set the contact information and have an Electronic Service Agent (ESA) configured to begin actively transmitting data to IBM.
With these configuration options in place, you can immediately cycle Collection Services to create historical data for any *MGTCOL objects in your configured library. To find this function quickly in IBM Navigator for i, type Cycle in the Search Task field on the left navigation panel and pause.
Figure 5. IBM Navigator for i Search Task field
The search shows a list of options with matching text. Select Cycle Collection Services and press Enter to invoke the action.
Click OK in the Cycle Collection Services panel.
Figure 6. Cycle Collection Services panel
This provides you with a starting set of data to view using the Graph History function.
Creating historical data from existing collections
When you turn on historical data creation, the next cycle of Collection Services will cause all available data in *MGTCOL objects within the configured library to be used to create the Historical Data collection. You can also manually generate historical data from other *MGTCOL objects by using the Create historical data action in Manage Collections.
Figure 7. Create historical data manually
It is also available on the left navigator under Performance → All Tasks → Collections.
Figure 8. Create Historical Data task
Visualizing historical data with the Graph History function
After you have historical data created, the new Graph History function in the Performance task on IBM Navigator for i allows you to view this historical system performance data. As mentioned earlier, this enables you to view data over long periods of time, from days and weeks to months and years. The Graph History function provides access to the new Historical Data collection, both summary and detail performance data.
Graph History Summary charts
Historical Summary charts utilize the historical summary data to provide single metric charts to view and analyze performance over the selected time span. The metrics provided in the summary charts focus on the system monitoring data saved into historical data. From the Summary chart, the top contributor information from the historical detail data for that metric is available by selecting an interval. From that Top Contributor chart, the property data for a top contributing element is available by selecting the element.
To view the historical data, use the new Graph History task section in Performance. To start, select Summary under Performance → Graph History.
Figure 9. Summary action
The majority of the new function for the 7.3 release is in the Summary charts. There is one Detail chart available now, and more are expected to be made available in the future.
Figure 10. Historical Summary chart
The Historical Summary chart in Figure 10 shows the CPU utilization (average) for 4 days with 1 hour intervals.
To maximize the screen space for charting, click the Hide Navigation icon to close the left navigation panel.
Figure 11. Hide Navigator icon
When the chart is first displayed, it displays the data for the past month at one hour intervals. There are too many points to clearly distinguish each interval on the chart, so it is shown as a line. As you zoom in, the individual data points will be shown.
To zoom in a specific point or time interval, you can use the slider or the mouse wheel.
To use the mouse wheel, click the data point on the graph for the interval area that you want to zoom in, then use the mouse scroll wheel to zoom in.
Figure 12. Slider
To return to the chart as it was originally presented, click Reset.
Figure 13. Tooltips
As you move the mouse pointer over the plotted points on the graph, tooltips appear with the value charted and other detailed metrics about each interval data point.
Figure 14. Icons
The five icons at the top-right corner of the Summary chart allow you to modify the presentation of the panel. The options are as follows:
Chart only – Select the chart icon to view the data on the chart. This is the default.
Chart & Table – Select the combined chart and table icon to split the screen between the graph and the table.
Table only – Select the table icon to show the dataset only in table format.
Display SQL – Select the SQL icon to display the query used to obtain the data set charted on the graph. This is shown as a pop-up panel and the chart panel will not be affected.
Help – Select the '?' icon for help text.
Figure 15. Maximum graph value
The dial at the top of the chart allows you to zoom in the chart by lowering the graph range (maximum graph value) shown on the chart. Points that occur above this setting will not be seen on the chart.
When the Summary chart is first displayed, it shows data starting from today's date in the previous month up to today. The Context panel allows you to select any other available metric to be displayed.
Figure 16. Summary chart
The Context panel is closed when the chart is displayed or updated to provide more room for the chart. Expand Context to open it.
Figure 17. Context panel
Use the Context panel to view and modify the chart specifics such as metric, report dates, graph interval, and dates.
After any changes to the Context panel, click Refresh to update the chart with the new settings.
The Metric drop-down list enables you to switch to any of the other available Historical Summary charts. See Summary Metrics sidebar for the list of metrics. Selecting another metric to display will preserve other settings of the Context section such as graph interval and dates.
The Collection Type is *HSTFILE.
The Library defaults to QPFRHIST. This can be changed to another library where you might have copied or moved a Historical Data collection.
The Report Dates field determines the time span to be displayed on the graph.
The Graph Interval values available are based on the Report Dates range. For a larger range of time, the graph interval will be larger. For a smaller range of time, the graph intervals available will be smaller. For example, if one month of data is requested, you will be able to set the graph interval to 1 hour, 6 hours, 12 hours, or 1 day (24 hours). If a report date greater than one year is chosen, then the graph interval values available will be one month or one year. Table 3 shows the Graph Interval settings available based on the selected Report Date range.
Table 3. Report dates and graph interval settings
(range of time selected)
|Graph intervals available|
|Today||5 min, 15 min, 30 min, 1 hour|
|1 week||30 min, 1 hour, 6 hours, 12 hours|
|1 month (default)||1 hour, 6 hours, 12 hours, 1 day|
|1 year||6 hours, 12 hours, 1 day, 1 week, 1 month|
|Custom < 6 hours||5 min, 15 min, 30 min, View All Points|
|Custom > 1 year||1 month, 1 year|
The From Date and Time fields and To Date and Time fields allow you to select a custom range (report dates) to be displayed.
The Historical Summary chart initially displays data starting from one month ago and ending with the collection last ended (often at the default midnight cycle time). To modify the dates and times displayed, first modify the Report Dates field or change the From Date and Time fields and To Date and Time fields. Then, select a Graph Interval value and click Refresh.
To view historical data collected today, you must manually generate historical data for the active Collection Services *MGTCOL collection by using the Create historical data action in Manage Collections.
Understanding the Historical Summary chart
The Summary chart is created for one metric over the specific time range selected in the Context section. Each point on the graph is shown with a symbol and color which indicates information about the data charted at that time interval.
A circle represents a data collection point that only has historical summary data available.
A square or triangle both represent data collection points that also have historical detail data available. You can click on these to see the top contributors data for that interval. The Top Contributors chart appears at the right side of the Graph History window.
The outline of the data point symbol also indicates information about the data being charted. The data point may be collected, aggregated, or extended data point.
- A collected data point is a value taken directly from a historical database file.
Figure 18. Collected data point
The historical data requested interval and graph requested interval match at the highlighted data point.
- An aggregated data point is a calculated value from the metric
at multiple intervals in the historical database files. Aggregation is done when
the collected interval is less than the requested graph interval. Calculations
such as sum, average, or max are performed depending on the metric and what it
For example, if data is collected at 15-minute intervals and the chart is showing 1 hour graph intervals, the data point will be a calculation of metrics from 4 time intervals.
Figure 19. Aggregated data point
The historical data requested interval is smaller than the graph requested interval at the highlighted data point.
- An extended data point is a value for that time interval which
is not available in the historical database file. The data value is extended
from the last available interval value. This is needed to provide integrity to
the chart for the graph interval requested.
For example, if data is collected at 30-minute intervals and the chart is showing 5-minute graph intervals, the data points between the collected intervals will be extended.
Figure 20. Extended data point
The graph request interval is smaller than the historical data creation requested interval at the highlighted data point.
Top Contributors chart
To view top contributor information for an interval, the detail data needs to be available. This is indicated by a square or triangle plot point. Select the data point. The Top Contributors chart is displayed at the right side of the panel.
Figure 21. Summary chart with the Top Contributor panel
Figure 22. Top Contributors chart
This chart shows the top contributors for the temporary storage in the selected time period. A top contributor will be listed multiple times if it was a top contributor for more than one interval of an aggregated data point. (See aggregated data point.)
Figure 23. Top contributors with duplicate element
For example, in Figure 23, the triangle data points are those where the data was collected at 30-minute intervals and a 1-hour graph interval was requested. There are several top contributors that are duplicated because they happen to be top contributors for both intervals. For an aggregated data point, both values are included in the calculation for this specific requested interval. Note that the time stamp on the Top Contributors chart helps to differentiate them.
To resize this section, click the Hide Top Contributors button.
Figure 24. Hide Top Contributors
From the Top Contributors graph, you can view property data for a job or task, memory pool, disk unit, or communications line by selecting a bar from the Top Contributors graph.
Figure 25. Summary chart with Properties panel
The Properties panel appears below the Top Contributors chart.
Figure 26. Properties panel
The Properties panel contains all the information available for the job or task, communications line, memory pool, or disk pool. You can scroll down the panel to see all the fields.
The Actions menu allows you to open the Historical Detail charts. At this time, there is only one Detail chart (CPU Utilization Overview) available. More are expected to be created in future releases.
Figure 27. Actions menu
Graph History Detail charts
The Historical Detail charts provide a view of additional data beyond the single metrics provided in the Summary charts. The Detail chart utilizes multiple summary metrics to provide more information at each interval. From the Detail chart, the top contributor information for one metric is available by selecting an interval. From that Top Contributor chart, the property data for a top contributing element is available by selecting the element.
The CPU Utilization Overview Historical Detail chart provides the Partition CPU Utilization and CPU time used from the summary data. When drilling down to top contributors, the metric is CPU time charged to the job.
With the first release of Graph History, the focus was placed on the Summary charts. There is one Detail chart available now, and more are expected to be available in the future.
Historical Data and the Graph History analysis tool is included with the base operating system in IBM i 7.3.
For best results, make sure you have the latest level Group PTF for IBM HTTP Server for i installed on your system:
- IBM i 7.3 HTTP group SF99722 level 1 at GA.
- Refer to IBM System i Support to find the latest level available for 7.3.
User profile authority requirements
To access the historical data, a user profile needs to have *ALLOBJ authority or be included in the data authority list QPMCCDATA for access to the Collection Services data.
To use the performance commands in IBM Navigator for i (Configure and Cycle Collection Services), a user profile needs to have *JOBCTL authority or be included in the function authority list QPMCCFCN to access the Performance Tools functions.
We can't predict the future. We can't tell you today what your system performance will be tomorrow. However, with this new Graph History function, you can look back in time and better analyze your system performance over a long term. This will allow you to anticipate future performance. Turn on historical data now and in the future you can examine past system performance. This will give you hindsight about what is happening on your system today.
This 7.3 release with Graph History support and the new Visualize Monitor Data, which charts all metrics for a system monitor on one panel, provides another significant step in the replacement of Management Central function with IBM Navigator for i for a single system.
- Refer to IBM i Technology updates on developerWorks for Performance on the web.
- Refer to Knowledge Center for IBM i 7.3 – Performance.
- Follow Dawn May's blog, i Can… Investigate Performance Data and look for the latest blogs in the Performance category.