6. VACUUM Processing
Vacuum processing is a maintenance process that facilitates the persistent operation of PostgreSQL. Its two main tasks are removing dead tuples and the freezing transaction ids, both of which are briefly mentioned in Section 5.10.
To remove dead tuples, vacuum processing provides two modes, namely Concurrent VACUUM and Full VACUUM. Concurrent VACUUM, often simply called VACUUM, removes dead tuples for each page of the table file, and other transactions can read the table while this process is running. In contrast, Full VACUUM removes dead tuples and defragments live tuples in the whole file, and other transactions cannot access tables while Full VACUUM is running.
Despite the fact that vacuum processing is essential for PostgreSQL, improving its functionality has been slow compared to other functions. For example, until version 8.0, this process had to be executed manually (with the psql utility or using the cron daemon). It was automated in 2005 when the autovacuum daemon was implemented.
Since vacuum processing involves scanning whole tables, it is a costly process. In version 8.4 (2009), the Visibility Map (VM) was introduced to improve the efficiency of removing dead tuples. In version 9.6 (2016), the freeze process was improved by enhancing the VM.