1. The 86,400 Seconds Rule
There are 86,400 seconds in a day. In interview math, round this to 100,000 seconds for rapid mental arithmetic.
If a system handles 100 million requests per day: $100{,}000{,}000 / 100{,}000 = 1{,}000 \text{ QPS average}$. Assume peak traffic is $2\times$ to $3\times$ average ($pprox 2{,}500 \text{ QPS peak}$).
If a system handles 1 billion requests per day: $1{,}000{,}000{,}000 / 100{,}000 = 10{,}000 \text{ QPS average}$ (peak ~25,000 QPS).
2. Latency Numbers Every Engineer Must Know
Jeff Dean's classic latency hierarchy provides immediate sanity checks for your architecture:
- L1 cache reference: ~0.5 ns
- L2 cache reference: ~7 ns
- Main memory (RAM) reference: ~100 ns
- Read 1MB sequentially from memory: ~3,000 ns (3 µs)
- Solid-State Disk (NVMe SSD) random read: ~10–100 µs
- Read 1MB sequentially from SSD: ~1,000 µs (1 ms)
- Disk seek (rotational HDD): ~10 ms
- Datacenter roundtrip (same building): ~0.5 ms
- Cross-country roundtrip (SF to NY): ~40–60 ms
- Transatlantic roundtrip (NY to London): ~70–80 ms
3. Storage & Bandwidth Sizing Formulas
Storage: Multiply Daily Active Users $\times$ Events per Day $\times$ Average Payload Size $\times$ Retention Days $\times 1.5$ (indexes and overhead).
Bandwidth: Multiply QPS by average payload size. A service handling 5,000 QPS with 200KB payloads requires $5{,}000 \times 200\text{KB} = 1\text{GB/s} = 8\text{Gbps}$ network ingress.
Model this architecture on the HLDForge Canvas
Turn static box diagrams into living, collaborative systems. Drag and drop microservices, Kafka clusters, and cache tiers. Run automated SPOF detection and STRIDE threat modeling in real time.
Written by HLDForge Architecture Team
Staff Systems Architect
The HLDForge Architecture Team publishes practical engineering guides, real-world case studies, and high-concurrency blueprints to help engineers build scalable and fault-tolerant software.