Storage Architecture and Design
What this domain tests
Storage Architecture and Design covers the distributed storage fabric: replication factor, erasure coding, capacity planning, failure domains, I/O path, and performance isolation. The exam expects you to size usable capacity and predict behavior during failures.
Core rules
- RF2 keeps two copies and tolerates one node or disk failure; RF3 keeps three and tolerates two. RF3 costs more raw capacity but meets stricter availability.
- Usable capacity is raw capacity divided by the replication factor before metadata, snapshots, and reserve. Plan for snapshots and rebuild headroom.
- Erasure coding reduces capacity overhead for suitable cold data, but it depends on cluster size and failure domains. Do not apply it blindly to latency-sensitive workloads.
- Check both capacity and IOPS: thin provisioning can overcommit space, and snapshots consume capacity as they age.
Worked example
Six nodes provide 12 TB raw each, or 72 TB. With RF2, nominal usable is about 36 TB. Reserve 10 percent for metadata and rebuild, leaving about 32 TB before snapshots. If the requirement is to survive two simultaneous failures, RF3 gives about 24 TB nominal, about 21.6 TB after the same reserve. Choose RF3 only if the availability requirement justifies the capacity loss.
Traps
Confusing raw and usable capacity; ignoring reserve and snapshot growth; assuming RF2 survives two failures; enabling erasure coding on small clusters; and designing for average IOPS while missing peak latency.