CDN request pricing: estimate $ per 10k / 1M requests (and when it dominates)

When people search for "CDN cost", they focus on bandwidth ($/GB). But for many workloads, request fees are a real line item. The key is to model request fees as a separate bill component and keep units explicit.

0) Define "billable request" for your CDN

Each CDN defines billable requests slightly differently (protocols, classes, regions). For planning, treat a billable request as "an HTTP/HTTPS request handled at the edge", and then adjust if your provider splits classes.

• Some providers price HTTPS differently from HTTP.
• Some providers have regional tiers or different request classes.
• Cache hit rate changes origin traffic, but request fees are usually counted at the edge.

1) Estimate requests per month

Use the most direct source you have. If you already run a CDN, analytics is the best input.

1. From CDN analytics: use the provider's 30-day request count (best).
2. From RPS: convert request rate to requests/month, then model baseline + peak. Tool: RPS to monthly requests.
3. From logs: sample access logs and extrapolate carefully, then replace with analytics once available.

Do not ignore bots, health checks, and retries. Those are often the "hidden request volume" that makes request fees dominate.

2) Apply per-10k vs per-1M pricing (keep units explicit)

Two common formats:

• Per 10k requests: request fees = (requests/month ÷ 10,000) × price per 10k
• Per 1M requests: request fees = (requests/month ÷ 1,000,000) × price per 1M

If you have a price per 1M and your calculator expects per 10k, divide the price by 100 (and multiply by 100 to go the other direction).

Tool: CDN request cost calculator.

3) When request fees matter most (traffic shapes)

• API traffic: high RPS, small payloads, lots of 200/304 responses.
• Highly cacheable sites: bandwidth goes down, request count stays high.
• Bot spikes: request volume jumps while bandwidth barely changes.
• Many small assets: sites that serve many tiny objects per page view.

4) Practical levers to reduce request fees

• Cache better: caching does not reduce edge request fees, but it can let you simplify APIs and reduce origin traffic and retries.
• Reduce background traffic: health checks, synthetic monitoring, and cron pings.
• Bot mitigation: block or rate-limit obvious bot traffic (especially on expensive endpoints).
• Batch and paginate: one request can replace many small ones for some API designs.

Worked example (sanity check)

• 100 RPS sustained is ~259M requests/month (100 × 86,400 × 30).
• At $0.75 per 1M, request fees are ~259 × 0.75 = $194/month.
• If a bot spike doubles RPS for 5 days, add that as a separate peak line item.

Related tools

Sources (examples of request-fee line items)

• AWS CloudFront pricing
• Azure Front Door pricing

Related guides

Related calculators

FAQ