Collecting performance data from a schedule

In Rational® Performance Tester, a schedule is a way of simulating load when testing an application. A schedule can be set up to run multiple instances of a test or to run multiple tests, all at once, to simulate the effect of having multiple end users using the application simultaneously. For information about setting up tests and schedules with Rational Performance Tester, see Creating tests and Representing Workloads.

Prerequisites:

First, you must specify that your tests and schedule should generate the ARM monitoring data that will be collected and analyzed:

  1. Ensure the tests involved in the schedule have ARM monitoring settings selected. Open the test in the Performance Test editor.
  2. Select the test element that exercises the part of the application that is experiencing the performance problem.
  3. On the Test Element Details page, select the Enable ARM monitoring check box.
  4. Repeat the previous two steps for each test element for which you want to gather data.

    Notes:

    • Click the Parent link to move to the Test Element Details page for the parent of the current test element.
    • If many tests have poor performance, go to a root test node and select Enable ARM monitoring for it; this will enable ARM monitoring for all of that node's children.
  5. Save the changes.
  6. Next, enable ARM monitoring for the schedule. Open the schedule in the Performance Schedule editor.
  7. On the Schedule Element Details page, select the Enable ARM check box.
  8. Collecting profiling data from all the users defined in your schedule will usually result in too much data for you to effectively analyze. Enable the Only sample information from a subset of users check box. It is recommended that you select Fixed number of users and specify one (1) user per user group, unless you have specific reasons for collecting data from multiple users.
  9. Save the changes.

Next, to collect performance data for an application running under the schedule, create a new profiling configuration, with the following specifics:

  1. Select J2EE Application with Performance Schedule as the configuration type.
  2. On the Schedule tab, select the schedule you want to run.
  3. On the Execution Results page, specify a location for the test results to be stored.
  4. On the Profiling page, on the Overview tab, select J2EE Performance Analysis.
  5. Edit the profiling set, if desired, by clicking the Edit button.
    1. On the Components page, select the types of J2EE components to collect data from.
    2. Do not set filters or sampling settings, as they are ineffectual on schedules.
    3. Click Finish.
  6. Click Profile to attach to the agent, start monitoring and run the schedule.

Tip: If you do not want to change any of the default settings for the profiling configuration (all component, no filters, no sampling), you can simply right-click on the schedule and select Profile > J2EE Application with Performance Schedule, instead of the above steps.

Note 1: Unlike in other data collection scenarios, when you click Profile for a J2EE Application with Performance Schedule configuration, both monitoring and the schedule are automatically started so that data collection is synchronized with the schedule starting.

Note 2: If you monitor a subset of users, it is possible that the schedule will appear to exercise a subset of the hosts involved in the application. Data will be collected only from those hosts that are exercised by the subset of users. This may result in what appears to be missing data from other hosts (hosts that were exercised by the non-monitored users). There is no way to control which specific users are monitored.

Note 3: Filters and samples work on root transactions. For collecting data from schedules, the root transaction takes place on the first ARM-instrumented element to be run. If the entire schedule is ARM-instrumented (that is, the Enable ARM Instrumentation check box is selected for the top-level schedule element), there will only be one root transaction, and so filtering and sampling will be ineffectual.

When the schedule has finished running, the test results and profiling data are available in the locations specified in the profiling configuration. Only data generated by the schedule is included in the results. Any other activity on the Web application is filtered out and not included in the performance data collected. For example, if other people are using the application at the same time that the schedule is running, the collected data does not include data resulting from their actions.

Once you have collected the performance data, you can begin analyzing it and diagnosing the problem. You can view the data using several views including statistics views and sequence diagrams of class and object interactions.

Related concepts

Distributed performance and problem analysis tools overview

Related tasks

Creating a profiling configuration for runtime problem determination

Customizing realtime profiling settings

Collecting performance data

Terms of use | Feedback
(C) Copyright IBM Corporation 2005. All Rights Reserved.