System Design

When to use:

• • Complex workflows with multiple steps
• • Need for async processing
• • Multiple consumers of same event
• • Building for scale

Trade-offs:

• → Harder to debug (distributed traces required)
• → Eventual consistency (not immediate)
• → Event schema versioning complexity

Connections:

→ Neural architecture (transformers process tokens in parallel)

When to use:

• • Different read/write performance requirements
• • Complex domain logic on writes
• • Need multiple views of same data

Trade-offs:

• → Increased complexity (two models instead of one)
• → Eventual consistency between read/write
• → More infrastructure to maintain

When to use:

• • Multi-service transactions
• • Can't use 2PC (two-phase commit)
• • Need business-level compensation

Trade-offs:

• → Complex error handling
• → Compensation logic can be tricky
• → Not all operations are compensatable

Example:

E-commerce order: Reserve inventory → Charge payment → Ship product. If shipping fails: Refund payment → Release inventory.