RAID6Parrays

RAID6Parrays is a hypothetical storage redundancy concept that extends traditional RAID 6 by introducing an additional parity layer, referred to as Parrays. The goal is to improve fault tolerance and data availability in multi-disk arrays, particularly in environments that require protection against multiple simultaneous disk failures.

Concept and architecture

In a RAID6Parrays configuration, data is striped across a set of disks and three parity streams are maintained: P, Q, and Parrays. P corresponds to the conventional RAID 6 parity, while Q provides a second independent parity stream. Parrays is a third parity layer derived from the same data blocks using a separate coding approach. The parity streams are distributed across the drives according to a predefined layout to balance load and preserve redundancy. The exact implementation details, including the coding method used for Parrays, can vary between proposals or vendor implementations.

Fault tolerance and rebuild

RAID6Parrays aims to tolerate more than two simultaneous disk failures under appropriate conditions, leveraging the additional Parrays parity to reconstruct data that would be unrecoverable under RAID 6 alone. Rebuild procedures involve using all three parity streams to restore failed disks. The level of resilience depends on the number of disks, the parity layout, and the specific coding techniques employed. In practice, triple-failure recovery may require more drives and careful management of parity distribution.

Performance and trade-offs

The added Parrays parity incurs higher write overhead, as multiple parity computations must be updated on each write. Read performance is typically similar to RAID 6, while rebuild times can be longer or shorter depending on the coding strategy and workload characteristics. Usable capacity is reduced compared to a non-parity-only configuration, with additional parity data occupying a portion of total storage.

Status and usage

RAID6Parrays remains a conceptual or experimental approach rather than an established standard. Implementations are not widely standardized, and support varies by vendor. It is primarily discussed in academic, lab, or specialized enterprise contexts where extended fault tolerance justifies the additional complexity and cost. See also RAID levels, parity, and erasure coding.