Archived logs are copies of online redo logs. They can be used to recover a database to point of failure or to another point in time after a backup has been restored. Archive logging is also required if online backups are desired.
Once a redo log file is filled, Oracle moves to the next log file. The archiver process (ARCH) copies the recently-filled log to an alternate location. If the archiver reads from a log on the same physical device as the current log being written, the sequential writes of the log writer are disrupted. If the log writer falls sufficiently behind, the database can stall (since a log file cannot be reused until it has been archived).
Therefore, it is important to optimize the performance of the archiver. Contention between the archiver and the log writer can be minimized by alternating redo logs over two devices. The redo log writer can then write to one device, while the archiver is reading from the other device. Since the archiver must be capable of writing at least as fast as the log writer, the archive destination should either be a dedicated device, or a dedicated set of disks in a RAID 0+1 (mirrored and striped) configuration.