Lesson 4 — Storage Media, Environments, and Database Organization
Storage media by workload
- Disk / SAN: persistent stable storage; SAN groups/manages disk arrays; lower-value data can be tiered to cheaper/slower media.
- In-memory: volatile memory for very fast access; durability/recovery mechanisms still matter.
- Columnar compression: repeated column values compress well and analytical scans can avoid reading unused columns.
- Flash / SSD: persistent storage with lower latency than traditional disk.
The exam question is usually what workload problem is being solved—speed, persistence, analytical scan I/O, capacity, or cost—not hardware engineering.
Environment promotion path
Development → Test / QA / UAT → Production
- Development: first controlled place to create/exercise changes and patches.
- Test/QA/UAT: formal functional, integration, acceptance, and performance validation. Performance testing should resemble Production closely enough for results to be meaningful.
- Production: mission-critical live processing; last stop; tightly controlled database changes.
- Sandbox: sits beside the promotion path for experimentation/POCs. It must be isolated and never write back into Production.
Production-derived data in lower environments can still be sensitive. Isolation does not eliminate masking, access, privacy, security, or retention requirements.
Database organization
Hierarchical
Tree structure: parent can have many children; each child has one parent. Efficient when the business structure is truly hierarchical; rigid otherwise.
Relational
Relations/tables; commonly schema-on-write and row-oriented; strong for structured operational transactions requiring frequent updates/consistency.
Multidimensional
Analytical organization across dimensions, common in DW/BI/cube use.
Temporal
Tracks time explicitly: - valid time = when fact is true in the real world; - transaction time = when the database considered/stored the fact as true; - bi-temporal tracks both.
Non-relational / NoSQL
May use schema-on-read and structures such as document/tree, graph/network, key-value, or column-family approaches; often selected for scale, availability, flexibility, or distributed workloads.
Row vs column orientation
- Row-oriented: many columns for a small number of records → OLTP-like whole-row transactions.
- Column-oriented: few columns across many records → OLAP-like scans/aggregations.
These taxonomies can overlap: a database may be relational and column-oriented. Do not force all labels into one mutually exclusive classification.
Specialized recognition
Spatial databases support geometric measures/functions; key-value stores retrieve by key; triplestores store subject–predicate–object triples; specialized graph/geospatial/time-series systems solve purpose-specific needs.
Source: pp. 176–185.