In the data cache, the database system keeps those data volume pages that have recently been accessed for reading or writing. The data cache is used jointly by all simultaneously active users.
The data cache is dynamically dimensioned by the database system, i.e. you cannot configure the size of the data cache directly, but can only influence it implicitly by configuring the I/O buffer cache. The database system takes the pages required for the data cache from the I/O buffer cache. If the other I/O buffer cache user, the Converter, grows in size, the database system decreases the size of the data cache, if necessary.
The size of the data cache has the following effect on the performance of the database system:
· A large data cache makes a high hit rate possible. We recommend a hit rate of at least 98%.
· If a data cache is too large, this can lead to operating system swapping with a very high level of I/O activity.
Data is usually written from the data cache to the volumes using the pager. If a user task finds no free pages available in the data cache, the user task will write the data from the data cache to the volumes.
In general, the tasks do not write directly from the data cache to the volumes; instead, they write from the data cache to an I/O queue first. Special threads then perform the actual I/O operations, such as writing to the volumes, asynchronously and in sufficiently large units.
See also:
Database Manager GUI, Displaying Cache Information
Database Manager CLI, Displaying Information About the Database Instance