Controlling Parallel Media Recovery

By default, Oracle uses parallel media recovery to improve performance of the roll forward phase of media recovery. In parallel media recovery, the database uses a "division of labor" approach to allocate different processes to different data blocks while rolling forward, thereby making the procedure more efficient. The number of processes used is derived from the CPU_COUNT initialization parameter, which by default is equal to the number of CPUs on the system. For example, if parallel recovery is performed on a system where CPU_COUNT is 4, and only one datafile is recovered, then four spawned processes read blocks from the datafile and apply redo.

Typically, recovery is I/O-bound on reads from and writes to data blocks. Parallelism at the block level may only help recovery performance if it increases total I/Os, for example, by bypassing operating system restrictions on asynchronous I/Os. Systems with efficient asynchronous I/O see little benefit from parallel media recovery.

To override the default behavior of performing parallel recovery, use the RECOVER with the NOPARALLEL option, or RECOVER PARALLEL 0.