Putting It All Together: Worked Example
What you'll learn: How to combine all estimation techniques into one complete, real-world calculation from user activity to infrastructure costs.
Let's design a simple photo-sharing app and estimate everything end-to-end.
Starting Assumptions
- Daily Active Users (DAU): 10 million
- Each user uploads: 2 photos/day
- Each photo size: ~500 KB
- Read:write ratio: 100:1 (users view 200 photos/day)
Step-by-Step Walkthrough
Traffic (QPS):
- Uploads: 10M users × 2 photos = 20M uploads/day
- Upload QPS: 20M / 100,000 seconds ≈ 200 writes/sec
- Peak multiplier (×5): 1,000 peak writes/sec
- Read QPS: 200 × 10 = 20,000 reads/sec (×5 peak = 100,000)
Storage:
- Daily: 20M photos × 500 KB = 10 TB/day
- 5-year projection: 10 TB × 365 × 5 ≈ 18 PB total
Bandwidth:
- Upload: 200 writes/sec × 500 KB = 100 MB/sec (~800 Mbps)
- Download: 20,000 reads/sec × 500 KB = 10 GB/sec (~80 Gbps)
Servers:
- If each server handles 1,000 QPS
- Read servers needed: 20,000 / 1,000 = 20 servers (40 for peak)
- Write servers: 2-5 servers (with redundancy)
Caching:
- Cache 20% of hot content: 0.2 × 18 PB ≈ 3.6 PB (impractical—cache metadata instead)
Costs (rough AWS example):
- Storage: 18 PB × 360K/month**
- Bandwidth: 80 Gbps outbound ≈ $200K/month
Notice how each number builds on the previous: users → requests → data size → servers → costs. This systematic approach prevents you from missing critical dimensions.
Key Takeaway: A complete estimation flows logically from user behavior through traffic patterns, storage needs, bandwidth requirements, server capacity, and ultimately to infrastructure costs—each step validating the others.