Databases Reference
In-Depth Information
In production environments, we usually check the disk controller-level
information to see if the same disk controller is handling both the
file systems. We distribute the data files based on disk controller
allocation—distributing them on different mount points managed by
the same disk controller brings no performance improvement.
The operations to be performed are similar to those executed when moving the online redo
log files, but, this time, we don't need to shut down the database. We can take the tablespace
offline, and move the desired data files to the new locations using O/S commands—we used
mv
in the example, inform the database about the new data file locations, as in step 17, and
then bring the tablespace back online.
By executing the all the steps until now, you can see the following output:
In step 19, there is a final tip about distributing Oracle files on machines where the Oracle
database shares resources with other applications. We need to keep separate disks for the
database to avoid possible issues and slowing down of simultaneous access to the disk by
other applications.
There's more...
When we discuss different disks, we obviously refer to different physical disks, possibly using
different controllers. It's important to know which process uses which type of database file.
Data files are written mostly by
DBWn
processes; the
CKPT
process reads and writes only data
file headers. Server processes read data from data files.
Redo log files are sequentially written by the
LGWR
process and read by
ARCn
processes, when
the database is in
ARCHIVELOG
mode. Archived redo logs are only written by
ARCn
processes,
under the same conditions.
The
CKPT
,
LGWR
, and
ARCn
processes can only read and write control files.