21.1 Enterprise Architecture Mental Model
Enterprise Datasphere architecture has five interacting planes.
| Plane | Purpose |
|---|---|
| Source | Systems that create authoritative data |
| Integration | Acquire, replicate, transform and federate |
| Semantic | Define reusable business meaning |
| Consumption | Analytics, applications and data products |
| Control | Security, governance, lifecycle, operations and cost |
Reference flow
Source systems -> acquisition -> harmonization -> semantic contract -> consumption
The control plane cuts across every stage.
System of record versus system of insight
A source system owns transactional truth. Datasphere should generally provide governed analytical truth, not become an accidental replacement for the transactional application.
Data product thinking
A reusable data asset should have:
- owner
- purpose
- business definition
- consumers
- quality expectations
- security classification
- freshness
- lineage
- support model
Architecture test
Ask:
- Who owns the source?
- Who owns the semantic meaning?
- Who consumes the data?
- Where is the data persisted?
- What happens when the source is unavailable?
- How is correctness proven?