A standard user-facing web application. The components are: DNS (IC1) → CDN (IC2) for static assets → Load Balancer (IC3) → API Gateway (IC4) → Application Servers (IC7) → Cache Server (IC9) → Relational Database (IC12). Background jobs via Background Worker (IC8) + Message Queue (IC13). Observability: Metrics (IC22) + Logs (IC23) + Traces (IC24).
The questions: Where is the SPOF? (Single-node database without replication.) Where is FM5 risk? (If the application server makes synchronous calls to an external API on the critical path.) Where is FM7 risk? (When the cache expires on a high-traffic key.)
Data Intelligence archetype. Event collection API → Message Queue (IC13) as buffer → Stream Processor (IC21) for real-time aggregations → Object Storage (IC17) for raw event archive → Batch Processor (IC20) for historical aggregations → Data Warehouse → Search Index (IC18) for dashboard queries. Metrics (IC22) alert on pipeline lag and error rates.
The questions: Where is FM8 risk? (Schema change in the event format without versioning.) Where is FM9 risk? (Late-arriving events corrupting time-window aggregations.) What is the freshness at each stage?
RAG architecture. User query → API Gateway (IC4) → Application Server (IC7) → Vector Database (IC19) for embedding retrieval → LLM inference (Application Server acting as orchestrator) → response. Background pipeline: Documents → Batch Processor (IC20) for chunking and embedding → Vector Database (IC19). Cache Server (IC9) for frequently repeated query embeddings.
The questions: Where is FM8 risk? (Embedding model version mismatch between ingestion and query time.) Where is FM11 risk? (No visibility into retrieval quality — which documents were retrieved for which queries.) What is the freshness lag between document update and vector index update?