You can use the appropriate features of your hardware, operating system, or database to optimize the availability of a database instance. Note the following recommendations:
· Locate every volume type on a different disk.
· In a production system, the log area should always be mirrored with hardware-based tools. Use RAID 5 or RAID 1 systems. Make sure that your RAID controller delivers good write performance, and that a cache is backed up to the disk even in a power outage.
· If you cannot use hardware-based mirroring, then you can set the log mode of the database instance so that it uses software-based mirroring.
· Use normal log backups. Do not allow the log area to be overwritten.
To make sure there is always enough space to write the log entries, make regular log backups.
· Always have redo log management activated.
· If you want to ensure a high standard of availability, we recommend using RAID-5 or RAID-1 configurations for the data area. A disk crash and change will then not affect the operation of the database, if the RAID system is able to perform a corresponding recovery procedure.
· In a production system on UNIX, data volumes and log volumes should be located on raw devices. Raw devices achieve better performance, and can be more securely controlled in the event of a system failure.
· When using fault-tolerant hardware, it is best to only use the same type of hardware when you want to extend the capacity. For example, only extend RAID-5 systems with RAID-5 systems and mirror disks with mirror disks.
See also: