RAID 10
RAID 1+0: pairs of mirrors striped. 4+ disks, half capacity used. Best balance of performance + redundancy + rebuild time. Standard for databases and high-throughput workloads. Tolerates 1 disk per mirror-pair failure.
Distinct from RAID 0+1 (stripe then mirror) which has worse failure tolerance. Throughput: ~Nx for reads, ~N/2x for writes. Rebuild: only the failed disk is cloned from its mirror — minutes instead of RAID 5/6's hours. Cost: 50% capacity overhead. Too expensive for large cold storage; perfect for 4-16 SSD database pool. Modern alternative: storage pools in ZFS, Btrfs, Ceph offer RAID 10-likes with snapshots + compression.