Buffer pool simulation

Buffer pool simulation helps you determine the right size for your buffer pools.

You need to set buffer pool sizes to minimize read I/O operations, which have a detrimental effect on performance. However, determining whether a buffer pool is the right size can be difficult. In general, making buffer pools larger reduces read I/O operations, but larger buffer pools require more real storage. In addition, larger buffer pool sizes do not always reduce the number of read I/O operations.

You can use simulated buffer pools to measure how many read I/O operations can be avoided if the buffer pool size is increased. You can do buffer pool simulations during normal operation of your Db2 database system. Production workloads are unaffected by simulation. If you do simulation as you run your production workloads, you collect realistic data about those workloads.

Simulated buffer pools require significantly less real storage and CPU overhead than real buffer pools, because only information about pages, rather than the pages themselves, are stored in the simulated buffer pools. In a data sharing environment, no directory entries are used for pages in the simulated group buffer pools.

The most suitable environment for buffer pool simulations is a buffer pool that gets a significant number of synchronous read I/O operations, and the pages in the buffer pool are likely to be referenced again. Buffer simulations are not as suitable for buffer pools that are used for objects whose pages are not likely to be referenced again, such as buffer pools that are used primarily for work files.

Restriction: Simulated buffer pools do not simulate changed pages. The size of the simulated buffer pool has no effect on write thresholds. If a virtual buffer pool contains a significant number of changed pages, the buffer pool accuracy of the simulation results might be affected.

Restriction: Although buffer pool sizes can cause access path changes, buffer pool simulations do not provide information on access path changes. When you change the size of a simulated buffer pool, no access paths are modified. To test whether a change in the buffer pool size results in access path changes, see Modeling a production environment on a test subsystem.