Skip to content
Library/Core Concepts
Scale estimation

Bandwidth Arithmetic

1 min read

Compute the network bandwidth your system actually needs — most designs miss this until it's too late.

Compute the network bandwidth your system actually needs — most designs miss this until it's too late.

How It Works

Bandwidth is the forgotten dimension of scale estimation. The formula: peak QPS (see Back-of-Envelope Math) × average payload size × any fan-out factor. A video streaming service serving 1 million concurrent viewers at 5 Mbps per stream needs 5 Tbps of egress bandwidth — that's a major cost line item, not a footnote. For APIs: 10,000 QPS × 10 KB per response = 100 MB/sec = 800 Mbps of outbound traffic, which saturates a small cloud VM's default network allocation. Common misses: underestimating log shipping, forgetting CDN fill traffic, ignoring cross-region replication bandwidth. In interviews, state bandwidth in the same breath as QPS and storage.

Real-World Example

Netflix built its own CDN (Open Connect) specifically because its bandwidth needs outgrew what commercial CDNs could serve economically. At peak hours, Netflix accounts for a substantial fraction of global downstream internet traffic, measured in dozens of Tbps. Any design discussion for a video product that ignores bandwidth misses the primary cost driver.

Test Yourself

Scenario: Your mobile app serves video thumbnails. 2M DAU, each user loads an average of 30 thumbnails per session. Each thumbnail is 50KB. Peak traffic is 3x the daily average. What is the peak egress bandwidth requirement, and what are the design implications?

Get notified when we launch

One email when the full practice product is live. No spam.