Skip to main contentSkip to user menuSkip to navigation

Rules of Thumb

Capacity, tail latency, and back-of-the-envelope math

Not Started
Loading...

System design is all about making intelligent trade-offs under constraints. These battle-tested rules of thumb help you size systems, estimate capacity, and set performance targets without getting lost in premature optimization. They're not precise formulas—they're engineering heuristics that work in 80% of cases.

The meta-rule: Start with these numbers, measure actual performance, then adjust. Good estimates beat perfect calculations that arrive too late.

⚡ Quick Decision

Start with Cache When:

  • • Read-heavy workload (100:1 ratio)
  • • Response time under 100ms needed
  • • Database queries are expensive

Scale Horizontally When:

  • • Single server hits 1000 QPS
  • • CPU/memory limits reached
  • • Need fault tolerance

Optimize Network When:

  • • Global user base
  • • Large payloads (>1MB)
  • • Mobile users (4G/5G)

💡 For implementation guides and code examples: See our technology deep dives: Redis, Nginx, Observability

Interactive Rule Explorer

Click through different categories to explore the key numbers every system designer should memorize.

📊

Capacity Planning

User load to infrastructure sizing

1M users25x
= 40K daily active
Typical engagement ratio
Read:Write ratio500x
= 100:1 to 1000:1
Cache-friendly workloads
1 server1000x
= ~1K QPS
Typical web server capacity
Cache hit50x
= 10-100x speedup
Memory vs database

🧠 Memory Palace

Visual mnemonics to remember the most critical numbers. These stick better than raw data.

🔥
100ms Rule
Under 100ms = Fire-fast response (instant feel)
⚖️
100:1 Read Rule
Reads vastly outweigh writes (scale for reading)
🏠
1K QPS House
One server = cozy house for ~1000 requests/sec

🎯 When to Apply These Rules

✅ Use These Rules When:

  • • Initial system design and capacity planning
  • • System design interviews and estimates
  • • Setting performance budgets and SLAs
  • • Quick feasibility checks
  • • Choosing between architectural options

⚠️ Don't Rely on Them When:

  • • Building safety-critical systems
  • • You have actual production metrics
  • • Extreme performance requirements
  • • Unusual access patterns or workloads
  • • Final capacity planning (always measure!)
Remember: These are starting points, not gospel. Every system is different. Start with these numbers, then measure, profile, and adjust based on your actual workload.
No quiz questions available
Quiz ID "rules-of-thumb" not found