Concatenating trace data from SMF and GTF data sets

The Collect Report Data (CRD) function of Buffer Pool Analyzer or an equivalent batch job are the recommended methods of making DB2® trace data available as input for activity reports and bpd files.

Other methods, like the Db2 trace facility, also provide usable Db2 trace data in standard SMF and GTF data sets.

If you intend to exploit SMF or GTF data, ensure that it contains the IFCIDs required by Buffer Pool Analyzer, as described in Determining what to collect. Otherwise some of the functionality that Buffer Pool Analyzer normally provides will be missing.

If you are accustomed to Db2 trace classes instead of individual IFCIDs, ensure that the appropriate trace classes, covering the required IFCIDs, are included in the data. Note that IFCID 198, which is required by Buffer Pool Analyzer for data of data type Detail, does not belong to any specific trace class, and corresponding data might therefore be missing in SMF or GTF data. If detail data is required in your SMF or GTF data, and if it is to be collected by means other than the Collect Report Data (CRD) function of Buffer Pool Analyzer, you can collect this data explicitly by means of the START TRACE command, as follows:

-START TRACE(PERFM) CLASS(30) IFCID(198) DEST(SMF)

This command uses the generic multi-purpose trace class 30, which has no predefined IFCIDs assigned to it. The required IFCID 198 is explicitly specified.

Even so collecting performance data through GTF or SMF is attractive especially for long-term collection and larger volumes, be aware that GTF, when the destination data set has filled up with trace data, proceeds recording data by overwriting the oldest data in the data set. Allocate a data set large enough to hold the expected amount of trace data for the collection period.

Note that SMF and GTF data do not provide data in the short record format, as it is required for the simulation function. Therefore, for simulations, you must collect trace data by means of the functions provided by Buffer Pool Analyzer. The other Buffer Pool Analyzer functions accept the standard record format, even so short is recommended in Determining what to collect for several other reasons.

Details about trace data in SMF and GTF data sets are described in the Db2 11 Administration Guide.

Assuming that the SMF or GTF data contains the necessary IFCIDs, you can use it alternatively or together with trace data being collected through Buffer Pool Analyzer. The JCL command stream in Creating activity reports and bpd files describes how input data sets are specified. You can concatenate multiple data sets with the INPUTDD statement to create one logical data set and continue processing the trace data as usual. The rules for concatenating data sets apply.

If you use DFSORT, see the z/OS® DFSORT Application Programming Guide for rules that apply to the concatenation of data sets.

If the trace data from other sources misses catalog information, you can use the Catalog only option of the CRD function (or the corresponding parameter in a batch job) to collect only catalog information (see Configuring a collect task, if required). Catalog data is used to enhance trace data that is collected through SMF or GTF. When you concatenate both data sets with the INPUTDD statement, the database identifiers (DATABASE) and object identifiers (OBJECT) from the SMF and GTF data are mapped to the actual database and object names.

Restriction: The possibilities outlined so far are provided for experienced and interested users wanting to use trace data from other sources. Their use requires detailed knowledge about trace data, involved IFCIDs, and tools. Because of endless variations and possible drawbacks, these options are formally not recommended and not supported in the current version of Buffer Pool Analyzer.