advanced
Traffic estimates
Estimate reads, writes, fan-out, peak load, QPS, bandwidth, and growth to size the first architecture.
Traffic estimates translate users and behavior into QPS, bandwidth, and peak multiples. Start from DAU or MAU, actions per user per day, and read/write split. Apply a peak factor (2–10×) for viral events or daily spikes.
Rough math is enough: 10M DAU × 20 reads/day ≈ 2.3k read QPS average; peaks may hit 20k+. Estimates justify caching, CDN use, database sharding, and queue buffering — they do not need spreadsheet precision in forty-five minutes.
On interviews: show back-of-envelope calculation on the board, state assumptions explicitly, and explain how a ten× error would change the design.
Common pitfalls: using average load only; forgetting fan-out (one write creates N feed entries); ignoring webhook or notification amplification.
The trade-off is flexibility versus complexity—know when the simpler path is enough.
Checklist:
- Estimate DAU/MAU and actions per user.
- Compute average and peak QPS for reads and writes.
- Account for fan-out multipliers.
- Connect estimates to scaling decisions.