System Design Cheat Sheet (Quick Reference)

🔵 Resource · Cheat Sheet · Bookmark this

This is your one-page System Design cheat sheet — every core concept, number, and decision rule distilled into something you can skim the night before an interview. Each line links to the full deep-dive on GyaanPost when you want more. Bookmark it; you'll come back.

The interview framework (6 steps)

The full walkthrough: how to approach a system design interview. Clarify scope → do the math → sketch the API/schema → draw boxes → deepen the hard parts → find and fix the bottleneck.

Numbers to memorise

• Latency: memory read ~100 ns · SSD ~100 µs · network round-trip within a datacenter ~0.5 ms · cross-continent ~150 ms. Full list: latency numbers every engineer should know.
• Capacity math: QPS = daily requests ÷ 86,400; storage = items × size × replication × retention. See capacity estimation.
• Availability: 99.9% = 8.7 h/yr down · 99.99% = 52 min · 99.999% = 5 min. Trade-offs in latency, throughput & availability.

Scaling toolkit (what to reach for)

• Too many reads? → cache, read replicas, a CDN.
• Too much data / write load? → shard the DB; use consistent hashing to spread keys.
• Spread traffic? → load balancing (and know the reverse-proxy vs LB difference).
• Decouple / smooth spikes? → message queues (Kafka).
• Slow lookups? → indexing; Bloom filters to skip misses.
• Protect the system? → rate limiting + retries, timeouts & circuit breakers.

Decision rules (the "which one?" table)

• SQL vs NoSQL: need transactions/joins/strong schema → SQL; need huge scale/flexible schema/high write throughput → NoSQL. Full guide: SQL vs NoSQL.
• Consistency vs availability (CAP): during a partition you pick one. Money/inventory → consistency; feeds/likes → availability. See consistency models.
• Strong vs eventual consistency: eventual is fine for most read-heavy social data; strong for balances, bookings, and distributed transactions (Saga).
• REST vs gRPC vs GraphQL: public/simple → REST; internal/low-latency → gRPC; flexible client queries → GraphQL. See API design.
• Real-time push: WebSockets vs long-polling vs SSE.

The building blocks (know one of each)

Load balancer · cache (distributed cache) · relational + NoSQL store (key-value store) · message queue · object store (S3/blob) · CDN · search index. Assemble these and you can design almost anything — see the worked URL shortener, news feed, and chat system.

Advanced (staff-level talking points)

Consensus (Raft/Paxos) for leader election & replicated logs · LSM-tree vs B-tree storage engines · microservices boundaries & trade-offs · horizontal vs vertical scaling.

What to read next

• 6-Week Study Plan — how to work through all of this
• Design Patterns Cheat Sheet — the LLD companion
• The complete System Design guide (hub) — every article, organised

← Back to the hub · Design patterns cheat sheet →