Traditional: Identifying resource issues by using the Performance Monitoring

Edit online

Traditional:
You can identify potential resource issues by monitoring the usage of resources such as CPU, heap, JDBC connections, and thread pools, displaying the real-time historical data, and then looking for a correlation with process instances and tasks. For offline analysis, you can export the data in the dashboard to a spreadsheet.

Before you begin

To view Performance Monitoring data, users must be assigned the WebSphere® Application Server administrative user role "Monitor".

About this task

The dashboard allows you to start the real-time performance monitoring, configure the monitoring sampling rate, sampling time frame (window), and target application servers. Then you can select different resources to display in the dashboard. If you have enabled the performance monitoring before a performance problem occurs, you can return to the performance monitoring dashboard within the length of the sampling window to view, analyze, or export the performance data for the time period that includes the period of the performance issue.

The performance monitoring captures data for the following resources:

Java™ Virtual Machine

The CPU Usage (in percent) of the JVM.
The JVM heap usage and total memory available.

JDBC connection pool of the Workflow Server database

The number of managed connections that are in the free pool.
The percentage of the managed connections pool that is used.

Thread pools

The number of concurrently active web container threads.
The number of concurrently active BPM Event Manager threads.

Process instance and task event statistics

The number of process instances started in each sampling time period.
The number of process instances completed in each sampling time period.
The number of tasks received in each sampling time period.
The number of tasks closed in each sampling time period.

Procedure

Log in to the Process Admin Console using a user ID that has the administrative user role Monitor.
In the Server Admin area of the console, click Performance > Monitoring.
If the monitoring is not activated, click the toggle switch to turn the monitoring on.
The Configure monitoring window opens.
1. For Sampling rate, use the drop-down to select the sampling resolution.
2. For Sampling time frame, use the drop-down to select how long the monitoring data is kept before being discarded.
  Important: Because the monitoring data is held in memory, combining a frequent sample rate with a long sampling time frame might cause memory problems.
3. Select the application server that you want to monitor.
4. Click OK to start monitoring.
  Tip: To change the monitoring settings, deactivate then reactivate the monitoring to reopen the Configure monitoring window.
When the performance monitoring is active, the dashboard shows the performance graph for the selected resource over a corresponding graph of the process instance and task event rates. You can perform the following actions:
- To update the graphs with the latest data, click Refresh.
- To display a different resource metric, use the drop-down to select a different metric.
- To zoom both graphs in on a time period of interest, click two data-points in the performance graph. A message box is displayed indicating that the view is zoomed. Close the message box to restore the view to display all available data.
- To identify where allocated resources are insufficient, look for resources that become 100% used.
- To identify possible causal relationships of resource issues, compare the graph that displays resource statistics against the graph that displays the process instance and task event statistics for the identical time period. For example, a sudden resource problem might be caused by one of the recently started process instances.
- To display more information about the process instances and tasks at any point in time, click on a data point in the process instance graph. You can click on the process instance name to view it in the Process Inspector, which might help identify processes and tasks that are the cause of resource issues.
- To export the performance data that is displayed in the current graphs for off-line analysis:
  - To export the performance data as an Excel file, select the Microsoft Excel 2007 checkbox before clicking Export.
  - To export the performance data in the default comma-separated value format, as a .csv file, check that the Microsoft Excel 2007 checkbox is not selected, then click Export.
  The exported file has the prefix Performance_Monitoring_Report_. The location of the exported file depends on your browser configuration, which might allow you to select the location or it might use a default location for downloads.
  Remember: The sampling time frame defines how long the data is held in memory before being discarded. This determines how long the data is available to be exported before it is lost.

Results

You have identified which resource allocations are insufficient or too generous, and also identify any processes and tasks that are unexpectedly resource hungry. Having this insight allows you to reallocate available resources more effectively and helps to focus development efforts to reduce the resources that are used by the processes and tasks that have the biggest impact on resource usage.

What to do next

When you no longer require the performance monitoring, turn it off by clicking the toggle switch.