The two data collection agents

When you set up a profiling configuration to start monitoring an application, you choose the type of data you want to collect and analyze by selecting a profiling set. Depending on the profiling set you choose, the data collection infrastructure uses one or both of two agents to collect the data. The agent used is selected automatically to correspond to your profiling settings.

The ARM agent is most useful, and will be selected automatically, in the following scenarios:

• When analyzing J2EE application performance and application failures especially for distributed applications where event correlation is essential.
• When profiling ARM-instrumented applications. For example, if you ARM-enable a database, you could see a transaction go from your application into the database, see all the processing in the database, and see the response come back out, all displayed in one sequence diagram.
• When using a shared server, and you want to profile only your own transactions while filtering out all other activity on the server.

The JVMPI agent is most useful, and will be selected automatically, in the following scenarios:

• When doing any memory analysis (for example, leak analysis).
• In order to see object interactions (for example, for the UML2 Object Interactions view). In most situations, class interaction analysis (available with the ARM agent) is enough to find problems, but occasionally you will need to do further profiling at the object level.
• When profiling non-J2EE Java™ applications, such as J2SE or J2ME applications.

Each agent offers its own set of data collection features, as highlighted in the following table: