Capacity Estimation in System Design
It is about predicting amount of resources your system needs*(servers, storage, bandwidth, etc.)* to handle expected traffic.
Key Concepts
- Traffic - How many people use your app? How often?
- Request Types -
- CPU-bound: Needs lots of processing (ex - video encoding)
- Memory-bound: Needs fast access to RAM (ex - caching)
- I/O-bound: Depends on disk or network speed (ex - uploading files)
- Storage Needs - How much data will be saved? (ex- images, videos, metadata)
- Bandwidth - How much data goes in and out each second?
- Peak Load - Plan for sudden spikes in activity.
Key Metrics to Estimate
| Metric | Description |
|---|---|
| DAU (Daily Active Users) | Number of unique users per day |
| QPS (Queries Per Second) | Number of requests handled per second |
| Storage Requirements | Data generated and stored per day/month/year |
| Bandwidth | Data transferred in/out of the system |
| Concurrency | Number of simultaneous users or requests |
| Response Time | Time taken to respond to a request |
| Error Rate | Percentage of failed or erroneous requests |
Standard Latency Numbers to Remember
| Operation | Latency |
|---|---|
| L1 cache reference | 0.5 ns |
| Branch mispredict | 5 ns |
| L2 cache reference | 7 ns |
| Main memory reference | 100 ns |
| SSD random read (4KB) | 150,000 ns = 150 µs |
| Read 1MB sequentially from memory | 250,000 ns = 250 µs |
| Round trip within same datacenter | 500,000 ns = 500 µs |
| Disk seek | 10,000,000 ns = 10 ms |
| Read 1 MB sequentially from the network | 10,000,000 ns = 10 ms |
| Read 1 MB sequentially from disk | 30,000,000 ns = 30 ms |
| Send packet CA → Netherlands → CA | 150,000,000 ns = 150 ms |
Example Estimation
Scenario: You’re building a photo-sharing app
Key Assumptions -
- DAU = 100,000 users with 20 requests/user
- Average Upload Size = 3MB
- Average Read Size = 1MB read
- Peak QPS → 100,000 * 20 = 2 million requests/day (23 requests/second). So we can assume that peak QPS can be roughly double ≈ 50
Resource Estimates -
| Category | Notes | Calculation | Estimate |
|---|---|---|---|
| Daily Requests | - | 100,000 DAU × 20 request | ~2 million/day |
| Bandwidth | Mostly reads at 1MB each; uploads ~3MB × 1 upload/user | 2M × 1MB (reads) + 100K × 3MB (uploads) | ~2.3 TB/day |
| Compute | Avg CPU time/request = 100ms then 1 CPU core can do 10 req/s | 50 QPS / capacity per core(10) | ~5 CPU cores |
| Storage Growth | Uploads only; 3MB/image → adjusted for compression (retained data) | 100,000 x 3MB = 300GB. After compression might be 16GB/day | ~500 GB/month |
| Write Load | Uploads, captions, likes/comments | - | ~20–30% of traffic |
Conclusion
Capacity estimation isn’t about perfect numbers it’s about directional accuracy. With a few smart assumptions and standard latency benchmarks, you can design systems that scale gracefully and perform reliably.