Building Blocks

Reusable components you assemble in most system designs.

Load balancer

Distributes traffic across instances. L4 (TCP, fast) vs L7 (HTTP-aware routing). Algorithms: round-robin, least-connections, IP-hash. Health checks remove bad instances.

Caching

Cut latency + DB load. Cache-aside, TTL + LRU, local vs distributed (Redis). See Caching.

Database scaling

• Replication — read scaling + HA (leader/follower).
• Sharding/partitioning — write scaling by shard key. See DB Scaling.

Message queue / stream

Decouple producers/consumers, absorb spikes, enable async & event-driven. Kafka (stream), RabbitMQ/SQS (queue). See Messaging.

API gateway

Single entry point: routing, auth, rate limiting, aggregation, TLS termination.

CDN

Cache static assets at edge locations near users; reduces origin load and latency.

Rate limiting

Protect services from overload/abuse. Algorithms: token bucket, leaky bucket, fixed/sliding window. Often centralized in the gateway or Redis.

Consistent hashing

Map keys → nodes on a ring so adding/removing a node reshuffles only ~1/N keys. Used by distributed caches/DBs.

Consensus / coordination

Leader election, distributed locks, config — via ZooKeeper / etcd / Raft.

ID generation

Unique IDs at scale: UUID, Snowflake (timestamp + machine + sequence), DB sequences.

Idempotency & retries

Idempotency keys make retries safe; combine with timeouts, backoff, and circuit breakers (see Scalability & Resilience).