As users modify data in Adaptive Server, only the transaction log is written to disk immediately, to ensure that given data or transactions can be recovered. The changed or “dirty” data and index pages stay in the data cache until one of these events causes them to be written to disk:
The checkpoint process wakes up, determines that the changed data and index pages for a particular database need to be written to disk, and writes out all the dirty pages in each cache used by the database.
The combination of the setting for recovery interval and the rate of data modifications on your server determine how often the checkpoint process writes changed pages to disk.
As pages move into the buffer wash area of the cache, dirty pages are automatically written to disk.
Adaptive Server has spare CPU cycles and disk I/O capacity between user transactions, and the housekeeper wash task uses this time to write dirty buffers to disk.
Recovery happens only on the default data cache.
A user issues a checkpoint command.
You can use the checkpoint to identify one or more databasess or use an all clause.
checkpoint [all | [dbname[, dbname[, dbname.....]]]
The combination of checkpoints, the housekeeper, and writes started at the wash marker has these benefits:
Many transactions may change a page in the cache or read the page in the cache, but only one physical write is performed.
Adaptive Server performs many physical writes at times when the I/O does not cause contention with user processes.
