One page
System Design cheat sheet
What to keep at hand. If a line is not clear, go back to the matching lesson in the course.
Open the course: System DesignNumbers worth knowing
| What | How |
|---|---|
| Read from the CPU L1 cache | ~1 ns |
| Read from main memory | ~100 ns |
| Random read from SSD | ~16 to 100 µs |
| Network within a data center (round trip) | ~0.5 ms |
| HDD seek | ~10 ms |
| Europe to the US and back | ~100 to 150 ms |
Estimation formulas
| What | How |
|---|---|
| Seconds in a day | 86 400 ≈ 10^5 |
| Average RPS | DAU × actions per day / 86,400 |
| Peak RPS | average × 2 or 3 |
| Storage per year | writes per day × size × 365 |
| 1M requests per day | ≈ 12 RPS |
| Cache memory | ≈ 20% of a day's data |
Availability
| What | How |
|---|---|
| 99% | ≈ 3.65 days of downtime per year |
| 99,9% | ≈ 8.8 hours per year |
| 99,99% | ≈ 53 minutes per year |
| 99,999% | ≈ 5 minutes per year |
Building blocks
| What | How |
|---|---|
| Spread requests | load balancer (L4 / L7) |
| Offload reads | cache (Redis) + read replicas |
| Split writes | sharding by key |
| Serve static files | CDN + object storage |
| Slow work | queue + workers |
| Events between services | Kafka |