Storage Architecture and Capacity Planning
Domain 1 - Storage Architecture and Capacity Planning
The blueprint does not publish a fixed domain list, so these sections follow the technical areas the expert-level exam consistently tests. This one tests whether you can turn a workload profile into a defensible OceanStor design: disk-domain layout, RAID 2.0+ protection level, hot-spare policy, tier composition, and usable-capacity math. Expect scenarios where two designs are technically valid and only one meets the stated SLA.
Core rules
- A disk domain pools disks. RAID 2.0+ slices them into fixed-size chunks and builds chunk groups, so a LUN is striped across many disks and rebuild is per chunk group rather than per whole disk.
- Protection level is a business decision: RAID 5 for capacity, RAID 6 for two-failure tolerance, RAID-TP when three concurrent failures must not lose data - it needs the highest minimum disk count and gives the lowest efficiency.
- Hot spare is reserved capacity inside the domain. It is not usable capacity, and a domain with no spare has no automatic rebuild target.
- SmartTier moves chunks among SSD, SAS and NL-SAS tiers from I/O statistics, so tiering needs at least two tiers and a migration window that avoids peak load.
- SmartThin separates provisioned from consumed capacity, but snapshots, clones and copies all consume from the same pool.
Worked example
25 x 8 TB NL-SAS disks = 200 TB raw. Reserve 2 disks as hot spare, leaving 184 TB. RAID-TP at 12D+3P gives 80% efficiency, so 147.2 TB. Subtract about 2% for system and metadata: roughly 144 TB usable. With 1.3 thin overcommit, maximum provisioned LUN capacity is about 187 TB, but the alarm threshold must track consumed capacity, not provisioned capacity.
Traps
Quoting raw capacity as usable; forgetting metadata and hot spare; treating snapshot space as free; placing every disk in one domain so there is no tier to promote hot data into; selecting RAID-TP on a domain too small to build the required chunk groups.