Operating image delivery on CloudFront and Lambda@Edge
An edge path that transforms images on the way out, carrying 253 million requests and 4.27 TB a month.
- CloudFront
- Lambda@Edge
- S3
Specification
- Organisation
- Nudge Healthcare
- Period
- 2023 – 2025
- Composition
- CloudFront, Lambda@Edge, S3 origin
- Role
- Delivery path operation, transform behaviour and cache verification
Context
Image delivery ran through an edge path that transforms on the way out. At this request volume, a transform result that fails to cache raises cost and latency at the same time.
Approach
At 253 million requests a month, a transform result that misses cache is simply cost. I tracked edge-hit behaviour against request count and transferred bytes.
I looked for segments where one image was being regenerated as several variants: when the cache key and the transform parameters disagree, the edge treats every request as a new variant.
Transfer and request count are different signals. Bytes rising while requests hold flat means more variants; both rising means more traffic. Only reading the two axes together tells you which.
System
Transform once, serve from cache.
CloudFront receives the request. A cached variant is served from the edge; only a miss runs Lambda@Edge against the S3 source.
Measured
4.27 TB/month
Image delivery infrastructure.
253M/month
Approximately 253 million.
Boundaries
- The 4.27 TB and 253M figures describe the operating scale of the image delivery infrastructure. They measure something different from the 417 → 235 GB/day in project 02, and the two are never presented as one achievement.
- This entry records scale operated, not an improvement delivered.