RAID types
JBOD combines drives into a storage pool without redundancy. Usable capacity is the sum of all drives, and different drive sizes can be combined.
RAID 0 stripes data across two or more drives for performance and capacity. It has no fault tolerance: a single drive failure loses the entire array.
RAID 1 mirrors data, usually across two drives. Read performance can improve, while writes are similar to a single drive. A two-drive mirror tolerates one drive failure.
RAID 5 uses distributed parity and requires at least three drives. It tolerates one drive failure. Rebuilding a failed drive can substantially reduce read and write performance.
RAID 6 uses dual distributed parity and requires at least four drives. It tolerates two drive failures, at the cost of additional parity overhead.
RAID 10 combines mirroring and striping, requiring at least four drives. Half the raw capacity is usable. Multiple failures can be tolerated provided no mirror group loses all its drives.
RAID 50 stripes across RAID 5 groups and requires at least six drives. It can tolerate one failed drive in each group. Failure tolerance depends on which groups contain the failed drives.
RAID 60 stripes across RAID 6 groups and requires at least eight drives. It can tolerate two failed drives per group. Losing more than two drives in any group can cause array failure.