Effects from concatenated input data sets

If you concatenate multiple data sets containing buffer pool performance data (no matter whether they were created by means of the Collect Report Data (CRD) function or by any other means), and if you use this data as input to create activity reports or bpd files, you should be aware of some side effects concerning the results.

• The input data sets you are concatenating might contain performance data from overlapping or segregative collection time frames.
• Summary information in activity reports is based on data that is collected at so-called statistics intervals (opposed to actual counts of events for detail reports). This means, a counter in a summary report is computed as the difference between the first and latest value covered during the collection of data.

Knowing this, it becomes obvious that summary information created from concatenated data can become imprecise or even useless because of various events.

• A DB2® system might be restarted between two collection time frames, which resets the counters being sampled at statistics intervals.
• The performance data in different data sets might be collected with different statistics interval settings or with overlapping time frames, which makes it impossible to compute valid results.
• Objects in buffer pools might be created or dropped during different collection time frames, or new objects might be assigned to buffer pools with previously used identifiers, which makes object related counter values invalid.
• Catalog information, correlation data, and aliases might be different and might not match the concatenated data, which also causes invalid results.

In summary, concatenating performance data to create summary activity reports is not recommended. This method, if used at all, is more suitable to detail activity reports, which are based on actual event counts in the concatenated data sets.

Preliminary remarks about the accuracy of summary and detail reports discloses further details about statistics-based and event-based data collection and the use of this data for activity reports.