Scenario · Last verified 2026-08-25
Video on demand
Origin library plus internet transfer for a modest VOD catalogue.
| Metric | Value | Unit | Source type | Derivation | Link |
|---|---|---|---|---|---|
| Library on origin | 2000 | GB | Author assumption assumption | Roughly 1,000 hours of 1080p mezzanine at about 2 GB per hour. Delivery renditions may live on a CDN origin instead. | — |
| Watched bytes | 9000 | GB | Author assumption assumption | 8,000 hours watched at 2.5 Mbps average (720p-ish). 8000 * 3600 * 2.5e6 / 8 / 1e9 = 9,000 GB. Peak live events are not included. | — |
| Ingest and publish | 20000 | request | Author assumption assumption | Multipart uploads for new titles and packaging outputs, on the order of tens of thousands of PUTs. | — |
| Segment GETs at origin | 8000000 | request | Author assumption assumption | HLS/DASH segments for 8,000 watched hours if origin sees every segment. A packager CDN should cut this sharply. | — |
Adjust the assumptions
| Vendor | Monthly total at the hub usage |
|---|---|
| Cloudflare | $29.85 |
| Azure | $36.71 |
| Google Cloud | $39.90 |
| AWS | $850.17 |
Breakeven: cdn-vs-direct-egress
Assumptions and limits
Nine terabytes of watched video is past every hyperscaler internet grant in this snapshot, so egress—not the two terabytes stored—sets the ranking. A media CDN or R2-as-origin is the decision, not a cheaper Standard class. Bitrate assumptions matter more than catalogue size: 2.5 Mbps times hours watched is the formula, and 1080p or 4K multiplies it. Segment GET counts look alarming until a CDN absorbs them; do not budget origin Class B as if every player hit S3. This model ignores DRM license servers and packaging clusters. Put the watched-gigabyte figure into the egress calculator first, then storage. Official transfer tables: https://aws.amazon.com/ec2/pricing/on-demand/ https://developers.cloudflare.com/r2/pricing/