Storage Concepts Index

Storage concepts, connected

Use this index when you need the right mental model quickly, then open the deep dive or operate the related lab.

Concept	The question it answers	Primary cost	Practice
Write Amplification	Why does one logical write become many physical writes?	endurance, bandwidth, latency	pipeline explorer
Write-Ahead Log	How can a mutation become durable before the main structure is updated?	sync latency, log capacity, recovery work	inspect the write path
Copy-on-Write	How can a system publish a new version without destroying the old one?	fragmentation, metadata writes, pinned space	reason about snapshots
Erasure Coding	How can data survive fragment loss with less capacity than replication?	fan-out, decode, repair traffic	failure lab
RAID levels	How do disk layouts trade capacity, fault tolerance, and rebuild behavior?	usable capacity and degraded risk	universal planner

One write through the stack

application mutation
  → WAL establishes recoverable intent
  → database/filesystem updates structures
  → CoW may publish a new version
  → RAID, replication, or erasure coding protects data
  → device firmware places and eventually reclaims media

The layers solve different problems and may coexist. Their costs compound, which is why StorageCraft keeps the measurement boundary visible.

Fast diagnostic route

1. Durability unclear? Start with the WAL acknowledgment and recovery contract.
2. Physical writes unexpectedly high? Trace amplification across every layer.
3. Snapshots consuming capacity? Inspect CoW reachability and fragmentation.
4. Protection looks efficient but recovery is slow? Model degraded reads and repair domains.
5. Choosing a layout? Compare usable capacity, tolerated failures, and rebuild exposure together.

Redundancy is not a backup, mathematical tolerance is not a placement policy, and an acknowledged write is only as durable as the complete flush and recovery contract.