|
Oracle advises that you distribute your database files in a way that facilitates performance and reliability. At least two active copies of the database's control files should be kept on at least two different devices. Redo log files should be multiplexed and each group of members be put on different disks. Data files whose data will have different life spans (as in a data warehouse or read-only tablespaces versus an Online Transaction Processing system) should be distributed on different disks. Data with different life spans (for example, application data and temporary sort data) should be separated into different tablespaces (and by extension into different datafiles) to minimize fragmentation. Data with different administrative characteristics (read-only tablespaces separated from highly active tables separated from marginally active tables) should be separated onto different disks. Any data, or data and indexes on that same data, that will participate in resource contention should be placed on different physical disks.
Oracle further suggests that data files be named with the root of the name relating to the context of the data that will be stored within that data file. For example, General Ledger data may have a GL prefix to the data file such as gldata or glindex, and user-specific data may have a prefix of u or user as in data or userdata. |
|