Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Architecture - Quick Revision

📌 Last-minute revision before interviews. Scan these points quickly.

Architecture Styles

  • Monolith: Single deployable, shared DB, simple but inflexible
  • SOA: Services share resources via ESB, coarser-grained
  • Microservices: Independent services, own data, complex but scalable
  • Serverless: FaaS (Lambda), pay-per-use, no server management

Scaling

  • Vertical: Bigger machine (simple, limited)
  • Horizontal: More machines (complex, unlimited)
  • Stateless: Easy horizontal scaling
  • Stateful: Need sticky sessions or shared state

Database Selection

  • SQL: Structured, ACID, joins, relationships
  • Document (MongoDB): Flexible schema, hierarchical
  • Key-Value (Redis): Simple lookups, caching
  • Column (Cassandra): Time-series, write-heavy, HA
  • Graph (Neo4j): Relationships, social networks

Communication

  • REST: HTTP, stateless, CRUD, public APIs
  • gRPC: HTTP/2 + protobuf, fast, typed, internal
  • WebSocket: Full-duplex, real-time, persistent
  • Message Queue: Async, decoupled (Kafka, RabbitMQ)
  • GraphQL: Client queries, single endpoint

Caching

  • Cache-Aside: App checks cache → DB on miss
  • Write-Through: Write to cache + DB simultaneously
  • Write-Behind: Write to cache → Async to DB
  • Eviction: LRU, LFU, TTL
  • Levels: Client → CDN → API Gateway → App → DB

Message Queues

  • Kafka: High-throughput, event streaming
  • RabbitMQ: Task queues, request-reply
  • Patterns: Point-to-point, Pub/Sub, Fan-out

Reliability Patterns

  • Circuit Breaker: CLOSED → OPEN → HALF-OPEN
  • Bulkhead: Isolate components
  • Retry + Backoff: Exponential delay
  • Rate Limiting: Throttle requests
  • Idempotency: Same operation, same result

Design Patterns

  • CQRS: Separate read/write models
  • Event Sourcing: Store events, not state
  • Saga: Distributed transactions via compensating actions
  • Strangler Fig: Incrementally replace monolith
  • Blue-Green: Two environments, instant switch
  • Canary: Gradual rollout

Distributed Systems

  • CAP: Consistency, Availability, Partition Tolerance (choose 2)
  • Consistency: Strong (always latest) vs Eventual (converge eventually)
  • Sharding: Horizontal partitioning (hash, range)
  • Replication: Master-slave, master-master
  • Consistent Hashing: Even distribution, minimal reshuffling
  • Consensus: Raft, Paxos

Deployment

  • Rolling: Update one by one
  • Blue-Green: Switch traffic
  • Canary: Route small % first
  • Feature Flag: Toggle without deployment

Observability

  • Logs: What happened (ELK, Loki)
  • Metrics: How much (Prometheus, Grafana)
  • Traces: Request flow (Jaeger, Zipkin)
  • Key metrics: Latency (p50/p95/p99), throughput, error rate, saturation

Security

  • AuthN: Who are you? (JWT, OAuth)
  • AuthZ: What can you do? (RBAC)
  • Encryption: In transit (TLS) + at rest (AES)
  • OWASP Top 10: Common vulnerabilities

Microservices Communication

  • Synchronous: REST, gRPC (request-response)
  • Asynchronous: Message queues, events
  • Service Mesh: Istio, Linkerd (sidecar proxy)
  • API Gateway: Centralized entry point

Key Concepts

  • Back-pressure: Slow down producers when consumers can’t keep up
  • Chaos Engineering: Inject failures (Chaos Monkey)
  • 12-Factor App: Cloud-native best practices
  • Blue-green vs Canary: Blue-green = instant switch, Canary = gradual

System Design Framework

1. Requirements (5 min): Functional + Non-functional + Capacity
2. High-Level Design (10 min): Components + Data flow + API
3. Deep Dive (20 min): DB schema + Caching + Scaling
4. Trade-offs (10 min): Pros/cons + Alternatives

🔗 Cross-References