Simulating buffer pool behavior

This function simulates different object placements and buffer pool size ranges in a representative performance snapshot and lets you see and interpret the effects. You can use it to perform what-if scenarios to balance buffer pool sizes and performance. It is performed on the client and uses buffer pool performance data (as raw DB2® trace data) as input. The file containing the trace data (in short format) needs to be downloaded to the client before it can be used.

If you use this function to test the recommendations from the object placement tool, the trace data file downloaded should be ideally the file from which the bpd file for object placement was created. To directly start the simulation function, right click the object placement result, then select Simulate....

The trace data that is used for simulations must be a good representation of the activity of the buffer pools that you want to optimize, as described in Determining when and how long to collect. This function uses recorded detailed activity of each object to accurately determine the effects of changes to the system.

You use the simulation tool on the client to select an appropriate trace data file and to perform simulations. The GUI provides an easy-to-use wizard that guides you through the few steps to adjust the simulation parameters and to assign objects to buffer pools. The wizard uses defaults that are based on information from the trace data file, but you can adjust these defaults interactively to see the effects of changes.

  • You can vary simulation parameters, such as:
    • The buffer pools to be included in a simulation.
    • The minimum and maximum buffer pool sizes to be simulated, and the increments by which the sizes are varied during a simulation.
    • The sequential steal thresholds to be simulated.
  • You can vary the assignment of Db2 table spaces and index spaces to buffer pools

Results from simulations are kept on the client. You can select them from the Buffer Pool Analyzer main window and view them in a web browser. The results show the recommended distribution of memory between buffer pools, and detailed information about misses and hit ratios as functions of buffer pool size.

You can also work with different performance scenarios by using different trace data files, and you can apply different simulation parameters to these simulations. By comparing the results, your predictions about the effects of changes become more reliable.

Simulating buffer pool behavior describes how to perform simulations and shows examples of simulation results.