AWS DEA-C01 ingestion guide covering sources, transforms, orchestration, code patterns, and loading decisions.
This is the heaviest DEA-C01 domain because most data-platform decisions start here. The exam wants you to distinguish streaming from batch, choose the right transformation engine, orchestrate multi-step workflows cleanly, and avoid fragile pipeline logic.
AWS currently weights Data Ingestion and Transformation at 34% of scored content.
flowchart LR
S["Source system"] --> C["Capture: batch, stream, API, or CDC"]
C --> L["Land raw data durably"]
L --> V["Validate, dedupe, and quarantine"]
V --> T["Transform into curated shape"]
T --> O["Orchestrate retries, alerts, and dependencies"]
O --> P["Publish to lake, warehouse, stream, or downstream app"]
For DEA-C01, ingestion is not just “move data.” A strong answer preserves raw evidence, makes replay possible, handles duplicates, validates schema, and leaves operators enough logs and metrics to explain failure.
| Lesson | Focus |
|---|---|
| 1.1 Ingestion Patterns, Sources & Triggers | Learn streaming versus batch ingestion, API consumption, triggers, schedulers, and replayability. |
| 1.2 Transformation Services, Formats & Processing Trade-Offs | Learn service selection for transformation, format conversion, integration, and cost-performance trade-offs. |
| 1.3 Orchestration, Workflows & Notifications | Learn workflow coordination, serverless orchestration, resilience, and alerting. |
| 1.4 Programming, IaC & Code Performance | Learn the coding, Lambda, IaC, CI/CD, and software-engineering concepts DEA-C01 expects data engineers to recognize. |
| Scenario clue | Strong first instinct | Weak answer pattern |
|---|---|---|
| Source database changes must feed analytics | CDC with restart position, raw landing, dedupe, and schema handling | Nightly full extracts forever without replay control |
| Records arrive continuously and need low latency | Stream ingestion with retention, ordering expectations, checkpoints, and consumer monitoring | Batch file polling with no latency guarantee |
| External SaaS or API is the source | Managed connector or bounded API job with pagination, rate limits, and retry control | One request that assumes all data arrived |
| Transformation must scale over large files | Glue, EMR, or Spark-style processing with partition/file-size awareness | Lambda for long-running high-volume transforms |
| Pipeline has dependencies and manual recovery pain | Step Functions, MWAA, EventBridge, retries, alerts, and run history | A cron script with no ownership or state |
The most common wrong answer is a one-step data move. DEA-C01 usually wants the platform shape: capture, durable landing, transformation, orchestration, validation, monitoring, and recovery.