Why this section
System design is architecture under load and failure: how a system behaves when there’s ten times more data, when a node dies, or when a message arrives twice. These twelve chapters cover the building blocks that appear in every real design discussion and every design interview.
Chapters (in learning order)
- 01 Requirements: functional vs non-functional requirements
- 02 Back-of-Envelope Estimation: quick capacity maths
- 03 Vertical vs Horizontal Scaling
- 04 Load Balancing and API Gateways
- 05 Caching: strategies and invalidation
- 06 Message Queues and Pub-Sub
- 07 Replication, Partitioning and Sharding
- 08 CAP Theorem and Consistency: consistency models
- 09 Idempotency, Retries and Timeouts
- 10 Blob Storage Patterns: large files
- 11 Rate Limiting
- 12 Monitoring, Logging and Tracing: observability basics
A design conversation, in order
flowchart LR R[Requirements] --> E[Estimate load<br/>& storage] E --> H[High-level design<br/>C4 container] H --> D[Deep dives:<br/>data, scaling, async] D --> F[Failure modes:<br/>retries, idempotency,<br/>limits] F --> O[Observability<br/>& cost]
Use this order in design reviews and interviews. Most bad designs come from skipping the first two steps.
The building blocks map
| Problem | Building block |
|---|---|
| Too much load for one machine | Horizontal scaling + load balancing |
| Repeated expensive reads | Caching |
| Spiky or slow work | Queues |
| Too much data for one database | Partitioning |
| Node failures | Replication + retries |
| Huge files | Blob storage patterns |
| Abusive or runaway clients | Rate limiting |
| ”What just happened?” | Observability |
Main resources
- Designing Data-Intensive Applications (Kleppmann): Ch. 1, 5, 6, 8, 9 map to these chapters
- Azure Cloud Design Patterns
- The System Design Primer: a free overview with exercises