Collecting data

DB2® performance data is made accessible through the Db2 Instrumentation Facility as Db2 trace data. Db2 tools and products, including Buffer Pool Analyzer, can get access to this trace data through the Instrumentation Facility Interface (IFI), and the Db2 command START TRACE can be used to record trace data.

A few hundred different types of trace records exist for different purposes; each is identified by an Instrumentation Facility Component ID (IFCID). Buffer Pool Analyzer collects only buffer pool related IFCIDs.

Different data collection options result in a varying performance overhead to a Db2 subsystem. Further, not all Buffer Pool Analyzer functions require the same set of trace data. Therefore, when you use Buffer Pool Analyzer to collect trace data, you can specify what trace data you want Buffer Pool Analyzer to collect, when to collect it, and how to collect it. This keeps any overhead minimized for a given task.

Trace data is collected on the host. You can use ISPF and the Collect Report Data (CRD) function of Buffer Pool Analyzer, or you can use a batch job to collect data. The CRD function provides a menu-driven interface to interactively configure and control this task. In a batch job you can use JCL to prepare and configure this task. Both methods provide the same results. The CRD function provides instant feedback about the collection process and status, but requires that you are logged on to TSO/E. The batch JCL lets you prepare the job offline and run it unattended, but requires that you verify the success of the job. The CRD function might be more appropriate for shorter, single tasks. Batch jobs are better for longer, repetitive, or scheduled tasks.

When a task is started by one or the other method, the Db2 command START TRACE is used to actually collect the data. However, you do not need to care about the trace command and its parameters. The command is performed invisible.