DB2 Version 10.1 for Linux, UNIX, and Windows

Using self-tuning memory in partitioned database environments

When self-tuning memory is enabled in partitioned database environments, there is a single database partition (known as the tuning partition) that monitors the memory configuration and propagates any configuration changes to all other database partitions to maintain a consistent configuration across all the participating database partitions.

The tuning partition is selected on the basis of several characteristics, such as the number of database partitions in the partition group and the number of buffer pools.

Starting the memory tuner in partitioned database environments

In a partitioned database environment, the memory tuner will start only if the database is activated by an explicit ACTIVATE DATABASE command, because self-tuning memory requires that all partitions be active.

Disabling self-tuning memory for a specific database partition

Enabling self-tuning memory in nonuniform environments

Ideally, data should be distributed evenly across all database partitions, and the workload that is run on each partition should have similar memory requirements. If the data distribution is skewed, so that one or more of your database partitions contain significantly more or less data than other database partitions, these anomalous database partitions should not be enabled for self tuning. The same is true if the memory requirements are skewed across the database partitions, which can happen, for example, if resource-intensive sorts are only performed on one partition, or if some database partitions are associated with different hardware and more available memory than others. Self tuning memory can still be enabled on some database partitions in this type of environment. To take advantage of self-tuning memory in environments with skew, identify a set of database partitions that have similar data and memory requirements and enable them for self tuning. Memory in the remaining partitions should be configured manually.