Metadata architecture
Centralized, distributed, hybrid and bi-directional metadata architectures, each with an interactive diagram you can stress-test.
1 min read
Metadata has a lifecycle with layers: creation/sourcing → storage → integration → delivery → usage → control/management.
- 01Creation / sourcing
- 02Storage
- 03Integration
- 04Delivery
- 05Usage
- 06Control / management
The four architectures
| Architecture | How it works | Key trait |
|---|---|---|
| Centralized | Single repository holding copies of Metadata from sources | High consistency; global search possible |
| Distributed | Single access point, no persistent repository; retrieves from sources in real time (via object request broker) | Always current; no manual entries possible |
| Hybrid | Metadata moves source→central repo, but repo only holds user-added/critical/standardized/manual items | Near-real-time + enhanced; doesn't improve availability |
| Bi-Directional | Metadata can change in any part (source, integration, UI); feedback coordinated back to source | Advanced; repo must manage source changes |
Explore each architecture
Single repository holding copies of Metadata from sources
Key traitHigh consistency; global search possible
- Metadata
- Manual entries
- Dashed links: batch replication
- Click a source to target it
Try it on ETL pipelines
What users see
- Databasesv1 · copy, up to date
- ETL pipelinesv1 · copy, up to date
- BI toolsv1 · copy, up to date
- Manual entries0 saved
What happened
- Pick an action above to stress-test the centralized architecture.
Pros
- High availability (independent of sources)
- Quick retrieval
- Resolved structures (unaffected by proprietary systems)
- Can be transformed/enhanced
Cons
- Complex replication of source changes
- Costly maintenance
- Custom extraction modules/middleware