Propose a cloud data ingestion and serving pattern for streaming and batch on your preferred cloud (AWS/Azure/GCP). Choose between CDC, event sourcing, or micro‑batch for upstream systems, justify partitioning/compaction, and show how you ensure idempotency, schema evolution (e.g., optional fields), and PII tokenization. Include cost controls (storage tiering, TTL, file size targets) and an incident you would simulate in chaos testing.