Questions occasionally arise concerning RAID levels in the configuration of NetApp storage used for Oracle databases and other enterprise applications. Many legacy Oracle best practices regarding storage array configuration contain warnings about using RAID mirroring and/or avoiding certain types of RAID. Although they raise valid points, these sources do not apply to RAID 4 and the NetApp RAID DP and RAID-TEC technologies used in ONTAP.

RAID 4, RAID 5, RAID 6, RAID DP, and RAID-TEC all use parity to ensure drive failure does not result in data loss. These RAID options offer much better storage utilization in comparison to mirroring, but most RAID implementations have a drawback that affects write operations. Completion of a write operation on other RAID implementations may require multiple drive reads to regenerate the parity data, a process commonly called the RAID penalty.

ONTAP, however, does not incur this RAID penalty. This is because of the integration of NetApp WAFL (Write Anywhere File Layout) with the RAID layer. Write operations are coalesced in RAM and prepared as a complete RAID stripe, including parity generation. ONTAP does not need to perform a read in order to complete a write, which means that ONTAP and WAFL avoid the RAID penalty. Performance for latency-critical operations, such as redo logging, is unimpeded, and random data-file writes do not incur any RAID penalty resulting from a need to regenerate parity.

With respect to statistical reliability, even RAID DP offers better protection than RAID mirroring. The primary problem is the demand made on drives during a RAID rebuild. With a mirrored RAID set, the risk of data loss from a drive failing while rebuilding to its partner in the RAID set is much greater than the risk of a triple- drive failure in a RAID DP set.