PostgreSQL exploits MVCC to enable high concurrent access to the underlying data, and this means that every transaction perceives a snapshot of the data while the system keeps different versions of the same tuples. Sooner or later, invalid tuples will be removed and the storage space will be reclaimed. On one hand, MVCC provides better concurrency, but on the other hand, it requires extra effort to reclaim the storage space once transactions no longer reference dead tuples. PostgreSQL provides VACUUM with this aim and also has a background process named autovacuum to periodically and non-invasively reclaim storage space and keep the system clean and healthy.
In order to improve I/O and reliability, PostgreSQL stores data in a journal written sequentially, the WAL. The WAL is split into segments, and at particular time intervals, named checkpoints, all the dirty data in memory is forced to a specified position in the storage and WAL segments are recycled.
In this chapter, you have...
