Storage Fundamentals and Architectures
This domain tests whether you understand why storage behaves as it does: media characteristics, RAID and RAID 2.0+, cache behaviour, the I/O path, and how block, file and object architectures differ.
Media and the I/O path
Hard disks are limited by seek time and rotational latency, so IOPS stay low while sequential throughput is high. SSDs remove mechanical delay, but introduce write amplification, garbage collection and endurance ratings (DWPD or TBW). Every read or write traverses host HBA → fabric → front-end port → controller → cache → back-end → disk, so a bottleneck at any hop appears as application latency.
RAID and RAID 2.0+
Classic RAID binds whole disks: RAID 5 tolerates one failure, RAID 6 two, RAID-TP (Huawei triple parity) three, RAID 10 mirrors stripes. Huawei RAID 2.0+ virtualizes disks into chunks, groups chunks from different disks into a CKG, and applies the RAID level across chunks. Rebuild therefore touches only the chunks that were on the failed disk instead of the whole drive, shortening reconstruction from hours to minutes; hot spare space is distributed across all member disks, and pools expand online.
Architectures
Block (SAN) presents LUNs to hosts. File (NAS) presents shared directories over NFS/SMB. Object presents buckets over HTTP/S3. Centralized dual-controller arrays scale up; scale-out systems add nodes for capacity and performance together. Active-active controllers both serve I/O; active-passive pairs fail over instead.
Worked example
Twelve 8 TB drives, one configured as hot spare, RAID 6 across the rest: usable = (12 − 1 − 2) × 8 TB = 72 TB. Deduct filesystem and metadata overhead when quoting. Because arrays are marketed in TB (10^12) while hosts report TiB (2^40), that 72 TB appears as roughly 65 TiB on the host - not a fault.
Traps
Counting the hot spare as usable capacity; assuming RAID 2.0+ rebuild speed equals classic RAID rebuild; sizing by raw instead of usable capacity; forgetting that write-back cache loses data without battery or capacitor protection; treating RAID as protection against deletion or ransomware.