Egress Pricing: The Silent Cloud Killer
Learn the principles of Cloud Bandwidth billing, HTTP Request charges, and why massive startups fail despite low user counts simply by serving high-resolution video incorrectly.
What is Data Egress?
Cloud providers (like Amazon Web Services or Google Cloud) have a notorious billing trap: Ingress is free. Egress is deadly. You can upload 500 Terabytes of video into an S3 bucket entirely for free. But the exact moment a user attempts to watch that video, AWS charges you roughly `$0.09` for every single Gigabyte that leaves their physical data center. If 1,000 users download a 1 GB file, your bill instantly jumps by $90.00.
The Two Billing Vectors
- 1. Bandwidth Weight (GB transfer): The raw physical heft of the data served. Video and 4K uncompressed images dominate this cost. A heavy React Javascript bundle also bleeds budgets here.
- 2. HTTP Request Counting: Generating millions of tiny 1 KB files. Even though the actual bandwidth weight is practically zero, the CDN provider must computationally burn CPU routing headers. They usually charge `$0.01` per $10,000$ HTTP requests. APIs and microscopic WebSocket polling hurt this metric.
Common Provider Architectures
- AWS CloudFront: Highly complex tiered billing. North America and Europe cost roughly `$0.085/GB`. However, if your traffic accidentally routes through South America or India edges, the cost can mysteriously spike to nearly `$0.17/GB`.
- Google Cloud CDN: Generally slightly cheaper baseline pricing globally and highly effective at routing through Google's pristine private undersea fiber, but charges heavily for cache-invalidation commands.
- Cloudflare: Famous for advertising "Free Bandwidth" and unmetered standard plans. However, if their automated algorithms catch you abusing the system by serving pure 100% video without HTML web traffic attached, they strictly disable the cache and force an Enterprise negotiation.
If you need to analyze how drastically a CDN can actually offset your origin database CPU load, run your arrays against our Cache Hit Ratio System.