Creating a Session

Before you begin

A remote-enabled project is required.

About this task

Sessions store and organize the results of several performance tuning Activities. A Session must be created before any Activities are created.

Procedure

  1. Make sure the Performance Tuning perspective is open.
  2. Right click the Sessions node in the Performance Explorer view and select New Session.

    The New Performance Tuning Session Wizard will open.

  3. Enter a name for the new Session.
  4. Select a remote C/C++ project.
  5. Select a remote context for the project, consisting of the build host and project location. This should be the context that is used to build the application, and is where the application's source code can be found (if available).
  6. If the build host will also be used to run the application then leave Same as build host checked. If the application will be run on a different host then uncheck the box and select the runtime host. If the desired runtime host is not listed then cancel the wizard, create a connection to the runtime host, and then open the New Performance Tuning Session Wizard again.
    Note: On Linux, runtime data collection must run as the root user, so if you choose a remote Linux connection it must be connected as root.
  7. Select a folder on the runtime host where temporary files will be stored during the performance run. These files will be automatically deleted after the performance run completes.
    1. Click the Use Default button to select the default location for the user account on the runtime machine.
    2. Click the Browse button to select a specific location.
  8. Click Next.
    1. If your runtime system is a local Linux system, you will be asked to select a root access command. On Linux, runtime data collection must be run as the root user. Choose or enter a graphical command that can authenticate that you have root privileges and run a command as root; different Linux distributions and desktops have different commands. A terminal command such as su or sudo is not sufficient, since the data collection is not performed in a terminal. If no other commands are available, you may be able to run sudo in a graphical terminal, for example by entering xterm -x sudo as your root access command, but the terminal command must wait for the command it runs to complete before it exits;gnome-terminal is not a suitable terminal command for this reason.
    2. Click Next

    The second page of the wizard is used to enter the executables and shared libraries that make up the application. This page is optional, however it is strongly recommended to enter this information because it will help produce useful feedback on how to increase the application's performance.

  9. Add executables by clicking the Add button next to the Executables list, or click Scan to search the project area for executables.

    Many applications consist of a single executable, however if your application consists of multiple executables then please enter them all.

  10. Add shared libraries by clicking the Add button next to the Shared Libraries list.

    Enter the shared libraries that are built along with the application for which you would like to collect performance data. It is not necessary to list 3rd party or system libraries.

  11. Click Next.

    If your application contains a mixture of C/C++ and Java code this page allows you to specify the percentage of time that the application usually spends executing Java code. This information will allow Performance Advisor to provide feedback on the Java portion of your application.

    If your application does not contain any Java code then click Next to skip to the next page.

    If your application does contain some Java code then either enter the percentage of Java in the text box or use the slider to select a value between 1 and 100.

    Enter the path to one or more JREs on the runtime system that may be used by the application. You can click the Scan button to automatically scan the system for JREs installed in typical locations. Alternatively you can add each JRE by clicking the Add... button and browsing to the folder on the remote host that contains the JRE. Feedback on the installed JREs can be found in the System Scorecard view after running an Activity.

    If you wish to performance analyze your Java code, a Java5 or newer IBM JRE is required. If you select a single IBM JRE here, Performance Advisor may be able to perform this analysis automatically, see Creating a Hotspot Detection Activity for more information.

  12. Click Next.

    It is recommended to begin performance tuning by running a System Scorecard Activity. This Activity will: 1) validate the runtime host by checking for the required system tools and 2) provide suggestions for configuration changes that can increase the application's performance.

  13. On Power systems only, select the minimum platform on which the application is expected to run. Select Auto if you are not sure what the minimum platform should be.
  14. On Power systems only, select the ideal platform. Recommendations will focus on performance tuning for this platform. Select Auto if you are not sure what the ideal platform should be.
  15. Click Finish. The newly created Session will appear in the Performance Explorer

Feedback