--> Specifying profiling criteria

Specifying profiling criteria

Profiling an application requires the selection of one or more analysis type under the Java Profiling data collector. Data collectors specify how data should get collected and analysis types indicate what type of data is to be collected. Filters ensure that only relevant details are channeled out to the views. Using filters is especially useful when speed and efficiency is critical: the less data there is, the less impact it has on the system, and the faster it can be collected.

Prerequisites:

 

Configuring the profiling criteria for an application is broken down into the following sections:

Selecting the relevant data collectors and analysis types

  1. In your profiling configuration, select the Profile Settings tab.

  2. The tree pane displays data collectors at root level and analysis types as the children of the data collector. Expand the 'Java Profiling' data collector and notice the analysis types that fall under it (for example: 'Execution Time Analysis', 'Probe Insertion', and so on...).
    Select the analysis types that you are interested in under the 'Java Profiling' data collector. The features supported for profiling Java 1.5 and Java 1.4 based java virtual machines are described below.

     

 

Profiling Java 5.0 (Java 1.5) and higher java virtual machine based applications

Java Profiler support for profiling Java 5.0 and higher JVMs is based on the Java Virtual Machine Tool Interface (JVMTI) introduced in J2SE 5.0, which replaces the earlier experimental Java Virtual Machine Profiler Interface (JVMPI). The Profiler uses a substantially different approach from the JVMPI-based Profiler. Instead of relying on VM-generated events for monitoring application behavior (such as method entry and exit events, object allocations and so on) it uses Byte Code Instrumentation (BCI) to inject small pieces of code which track the application execution. This technique allows for efficient monitoring of specific parts of the application, without imposing a noticeable overhead on other parts of the application that are not being monitored. This approach is easy to use and scales well for large applications.

Execution Profile of an Application

The Java profiler can be used to profile any Eclipse plug-ins, Eclipse RCP applications, J2EE applications, and Java applications. To profile an application select "Profile Configurations..." from the Profile drop down menu or the Run menu. This opens the profile launch configuration window.

 
Perform the following steps:

Memory (Object allocation) analysis

To perform an Object Allocation analysis or Memory analysis (Heap), perform the following steps:

Thread Profile of an Eclipse Application

To perform a thread profile of your application, perform the following steps:

Probe Insertion

To instrument your application using probes, perform the following steps:

Setting profiling options

You can set profiling options for the Java Profiling data collector or the selected analysis types, by highlighting the entity and clicking Edit Options.

Follow the steps below to set profiling options on the Java Profiling data collector:

  1. Highlight Java Profiling and click on Edit Options
  2. By default, all classes are profiled. You can change this on the Filter Set page by specifying the filter set to be applied to your profiling session. Under the Select a filter set list, click Add to add the desired filter set.
  3. You can also specify the filtering criteria manually. Under Contents of selected filter set, click Add to add packages for which to collect data. The Add Filter dialog opens. An asterisk * can be used as a wildcard in a pattern when specifying package or method names.
    Notes:
    1. When entering a pattern for the package or method name, the pattern must be specified in one of three ways:
      • A pattern with no wildcards, e.g. org.eclipse.tptp.
      • A pattern with a wildcard at the beginning, e.g. *.sample.myapplication.
      • A pattern with a wildcard at the end, e.g. org.eclipse.*.
    2. Certain classes will always be visible in the profiling views regardless of the filter settings. The profiling agent will always collect data for objects or classes that are associated with the task being examined while collecting data specific to the task. For example, we specify a filter to collect data for a single class MyUserClass, and specify another filter to exclude the java.lang.* classes. When an instance of MyUserClass is loaded though, it has an associated class definition that gets loaded. This class definition has an associated class object, which is in turn of type java.lang.Class. This results in data collection for the java.lang.Class in addition to the MyUserClass class.
      This behavior applies to threads as well. The java.lang.Thread class will always show up in the Profiling views because the profiling agent tracks all running threads, and therefore, collects data for objects or classes associated with those threads.
    3. Filter configuration determines the scope of the data collection. Package of target application should be applied in filter configuration. eg. mypkg* * INCLUDE. It is highly not recommended to use an all inclusive set as a filtering criteria. eg. '* * INCLUDE'.
  4. Specify the Package or Class, Method Name that you wish to filter. If you want see the profiling data for these packages, classes, methods, select Include from the Rule drop-down list. If to exclude these packages, classes, methods, from the profiling data click on Include under the Rule column for the package, and change it to Exclude.

  5. Click OK. The filter criterion is added to the contents list.

  6. Click Finish.
  7. To start the profiling agent automatically when the application is launched, check the Automatically start monitoring when application is launched check box. It is strongly recommended for large applications that are not required to be monitored during the initialization phase (for example launching an instance of the workbench), that you accept the default and leave the Automatically start monitoring when application is launched check box not selected. This will result in the profiling agent being started in pause mode so that data is not collected during the initialization stage of the application.

  8. Click Finish

  9. Click Apply to save the changes.

Similarly, you can set the options of an analysis type by highlighting it and selecting Edit Options.

Configuring the profiling data destination

You can specify the project, and the monitor to use for your profiling session. You can also write the profiling data to a file.

  1. In your profiling configuration, select the Destination tab.
  2. Specify the Profiling Project to use for your profiling session.
  3. Specify the Monitor to use.

  4. To save the profiling data to a file, select the Send profiling data to a file and specify the file name in the Profiling file input box.

  5. Click Apply to save the changes.

  6. Click Profile to start profiling your application.
Hint: When filtering, only the first applicable filter is applied. When you are specifying filters, ensure that you declare the most specific filter criteria first.

 


Related

Overview of the Java Profiler


Tasks

Profiling an application
Setting profiling preferences


 

(C) Copyright IBM Corporation 2000, 2012. All Rights Reserved.