||Oracle Tips by Burleson
Chapter 6 Oracle Tablespace and File
In order to equalize disk I/O, it may be
beneficial to spread data files for large databases across several
drives. This must be decided on a case-by-case basis, and is
database-specific. If there are several large tables that would
benefit from being spread across multiple disks, consider placing
them in their own tablespaces, then size the data files for the
tablespaces such that the data contained in the tables is spread.
For example, a single table that contains a gigabyte of data may
benefit from spreading the file across several platters.
This can be done in Oracle7 by creating a
table-specific tablespace on each platter that will hold the file,
with each tablespace a fraction of the total size of the table. For
instance, if the file will be spread across four drives, each data
file would be 250 megabytes in size. Then, when the table is
imported, it will be spread across the four drives.
The database treats the table as one contiguous
entity, but spreading the table across the available drives
increases I/O speed. Under Oracle8, Oracle8i, and Oracle9i, this can
be accomplished with table partitioning. Partitioning allows a
single table to be spread, by value range, across several files. Of
course, with RAID 1, RAID01/10, or RAID5, the spreading is done
quasi-automatically, so the only reason to partition is for use of
partition elimination or the benefits of parallel query and
independent partition maintenance.
Each tablespace should be
created with a default storage parameter that takes into account the
performance demands of the tables in the application. Estimate the
size requirements for the tables as they are created, and