System Design

Core patterns for designing scalable, reliable distributed systems. Covers the building blocks you need for both production engineering and system design interviews.

Files

File	Topics	Level
scaling.md	3-tier architecture, vertical vs horizontal, DB read replicas, sharding, resharding, consistent hashing, celebrity problem, fan-out, circuit breaker, backpressure	SDE-1/2
cap-pacelc.md	CAP theorem, CP vs AP, consistency models (linearizable→eventual), PACELC, quorum math, vector clocks, tunable consistency	SDE-2
rate-limiting.md	Fixed window, sliding window, token bucket, leaky bucket, Redis Lua implementation, distributed rate limiting, nginx, AWS API GW	SDE-1/2
async-patterns.md	Message queues, pub/sub, DLQ, Saga (choreography/orchestration), outbox pattern, CQRS, event sourcing, idempotency, backpressure	SDE-2
api-design.md	REST vs GraphQL vs gRPC, versioning, pagination (cursor/keyset), idempotency keys, API gateway, auth patterns, webhooks, OpenAPI	SDE-1/2
distributed-transactions.md	Dual-write problem, 2PC, Saga, outbox pattern, CDC/Debezium, distributed locking (Redlock), optimistic concurrency, TCC, Spanner/TrueTime commit-wait	SDE-2
file-transfer-storage.md	Presigned URLs, multipart/resumable upload (tus), chunking + content-addressable dedup, replication vs erasure coding, metadata service design, CDN delivery	SDE-2
realtime-chat.md	WebSocket/long-polling/SSE, connection registry + cross-server relay, delivery guarantees, presence, multi-device fan-out, MQTT, WebRTC calling	SDE-2
distributed-id-generation.md	UUID v4 vs ULID/UUIDv7, Twitter Snowflake bit layout, clock-drift handling, ticket servers, range/segment allocation	SDE-2
geospatial-services.md	Geohashing, quadtrees, S2 geometry, Redis GEO commands, ride-hailing matching, KNN vs radius search	SDE-2
probabilistic-data-structures.md	Bloom filter recap, HyperLogLog cardinality estimation (live simulator), Count-Min Sketch frequency estimation	SDE-2

Read order: scaling → cap-pacelc → rate-limiting → async-patterns → api-design → distributed-transactions → file-transfer-storage → realtime-chat → distributed-id-generation → geospatial-services → probabilistic-data-structures

Mental Model

User Request
    │
    ▼
CDN / Edge Cache  ──── (static assets, public API responses)
    │ miss
    ▼
Load Balancer (L7)
    │
    ▼
Stateless App Servers ──── Rate Limiter ──── Auth (JWT/OAuth2)
    │
    ├──► Cache (Redis) ──── cache-aside, TTL, eviction
    │
    ├──► Database (primary) ──── write path
    │       └── Read Replicas ──── read scaling
    │       └── Shards ──── write scaling
    │
    ├──► Message Queue ──── async jobs, fan-out, retry
    │       └── Workers
    │
    └──► External APIs ──── circuit breaker, timeout, retry

When to Apply Each Pattern

Symptom	Pattern
App servers are CPU-bound	Horizontal scale + LB
DB reads are slow	Read replicas + Redis cache
DB writes are slow	Vertical scale → sharding
Single key getting hammered	Celebrity problem → key splitting / L1 cache
Slow synchronous operations (email, resize)	Async queue + workers
Downstream service is flaky	Circuit breaker + exponential backoff
Need distributed transaction	Saga + outbox pattern (avoid 2PC)
Need to prevent duplicate processing	Idempotency keys + dedup table
API is getting hammered	Rate limiting (token bucket)
Strong consistency required	CP system + quorum reads
High availability over consistency	AP system + eventual consistency

System Design

System Design

Files

Mental Model

When to Apply Each Pattern

Pages in this section