| Team Fly | |
the database. DBWn writes changes to the data files to free up dirty buffers (buffers that have changed data in them) to allow more changes to occur at system checkpoint time. A checkpoint is a background event that ensures all of the changed data has been written to the data files.
Archive logs are the mechanism used to provide continuous availability to an Oracle database by allowing you to back up the database while it is running. This is known as a hot backup. They also allow you to recover databases by performing a ''roll-forward" of changes using redo logs up to either the current point in time or to a time that we specify the database should be rolled forward to. This cannot be done with online redo logs because they are written to sequentially and eventually wrap-around. When that occurs, the previous records that existed in the wrapped redo file are overwritten. By placing the database in archivelog mode, online redo logs can be written out to files to be kept indefinitely. The files can be named with sequential incrementing names so that we know the order they should be applied in. As with redo logs, archive logs can be multiplexed. In other words, more than one copy of each archive log can be written to different locations.
Datafiles are the low level structures that make up what you probably think of as a database. To put it simply, the tables and indexes that make up your applications are stored in tablespaces and each tablespace is created on one or more datafiles. A particular datafile will store data for one tablespace and a tablespace can contain many tables and indexes. Tables and indexes are a subset of a type of database object called a segment. Examples of segments are indexes, tables, rollback/undos, clusters, lobindexes, and lobsegments. A segment is made up of extents that are a collection of contiguous blocks. So, to put this into a single sentence, a tablespace is created on one or more datafiles and consists of one or more segments made up of one or more extents that are a contiguous collection of blocks. Got it? Good. It's an important set of relationships to understand.
From a backup and recovery point-of-view, it's important to be aware of the relationship between tables, tablespaces, and datafiles. Whenever you perform a backup where you are not backing up the entire database at the same time, it is important to consider any referential integrity constraints enforced by the backup or application that must be honored. You should employ a naming standard that makes it easy to determine the tablespace that a data file supports.
| Team Fly | |