Show TOC

 Performance Data Within the Request Response CycleLocate this document in the navigation structure

Use

The Web Dynpro Console supports you when you analyze performance problems encountered by users, especially those concerning delayed response times. Delayed response times for a query can be caused by:

  • Poor performance by the Web Dynpro framework
  • Problems with the client, e.g. browser
  • Poor performance by the Web Dynpro application
  • Unacceptable performance by the Web Dynpro model or backend
  • Overloaded J2EE Engine
  • Suboptimal setting of the J2EE Engine

Under the tree node Performance in the Web Dynpro Console you can display performance data for the different components of a query, such as the Web Dynpro framework, or measure the performance of the Client component.

Using the data, you can analyze the reasons for the performance problems of a Web Dynpro application: Time values that deviate considerably for certain methods indicate potential performance problems.

Prerequisites

You have administrator rights for the J2EE Engine or for SAP Enterprise Portal. If you do not have administrator rights, you can only access a small part of the monitoring information.

Features

As shown below, the elements under the tree node Performance in the Web Dynpro Console represent the components involved in the request-response cycle.

The times needed to execute methods and components are either displayed automatically for these components as under Web Dynpro Framework(Element 1) or can be triggered by the users themselves as under Client.

For example, the data for a  Web Dynpro application is only displayed if you set the JARM level when building and deploying the Web Dynpro project (see Measuring Performance for Web Dynpro Applications ).

In production you would normally deploy an application without setting this JARM level, as otherwise additional source text would be created when the Java classes are generated. This source text would be executed when the application is called, which could possibly prolong the processing times in an application. In general the additional time required to measure the methods is between 10 and 20 msec per method, and thus can be ignored with regard to performance.