Process in memory
Filter, reshape, validate, enrich, mask, classify, and route records without first writing the full intermediate data set to permanent storage.
Architectural deep dive
Zero-Landing Architecture
Transform, validate, enrich, and route data in motion. Persist raw or intermediate data when it serves a requirement—not because the integration engine requires a staging layer.
In-memory processing
Inline transformation
Optional persistence
Fewer handoffs
Landing is a design choice, not a prerequisite
Many integration stacks must first copy source data into a landing area, then start another process to transform it. DataZen can perform those operations while data is moving through one pipeline, reducing the number of storage layers and orchestration handoffs needed for operational integration.
Push work to database compute
When a database is the right execution engine, pipeline stages can use its query and transformation capabilities without making every flow warehouse-centric.
Persist with intent
Keep raw history, replayable change logs, audit evidence, or analytical layers when they provide business value. Zero-landing does not prohibit storage.
Reduce architectural dominoes
Fewer intermediate systems mean fewer schedules, credentials, schemas, copies, retention policies, and failure boundaries to coordinate.
A continuous flow from source to destination
- Acquire. Read from the source through its existing database, API, file, storage, or messaging interface.
- Prepare inline. Apply schema rules, transformations, validation, enrichment, and change capture as the data moves.
- Make decisions. Branch or route based on data values, external function results, or AI-agent responses.
- Deliver or persist. Write the final result directly to its destination, adding durable intermediate storage only where required.
Zero-required-landing is the precise model
DataZen itself still uses operational state and may create change logs when capture or replay is part of the design. The architectural distinction is that a permanent Bronze or staging copy is not mandatory before transformation and delivery can begin.
Land data when the landing layer has a job
- Keep it: when history, compliance, lineage, independent replay, or broad analytical reuse requires a durable copy.
- Skip it: when data is transient, the target is operational, latency matters, and the intermediate copy has no consumer.
- Use change logs: when multiple downstream readers need reliable delivery or replay without retaining every source snapshot.
- Mix models: route one prepared stream directly to an application while persisting another for analytics or audit.
Remove storage layers that exist only to satisfy tooling, while preserving the ones that satisfy an actual data requirement.
Back to DataZen