Evaluate Your CDN by Its Cache Misses

Serving a cached file over a well-connected edge is not the same challenge as fetching fresh content from origin. The discussion argues that CDN evaluation often underweights this distinction, even though cache misses can contribute to rebuffering and bitrate issues. Live video makes that origin-to-edge path especially important because there is little margin for error.

The proposed approach starts with placing CDN ingest as close to the customer’s origin as possible, with a stated target of one to two milliseconds where achievable. From there, multi-hop relays shorten transport feedback loops, while parallel fetches let the fastest request win. The stakes extend beyond an individual request: a slow origin fetch can affect everyone depending on that content at an edge.

Key Topics

• Cache-miss performance as a CDN evaluation criterion
• Streaming quality and live delivery’s limited margin for error
• Origin-adjacent ingest and TCP responsibility
• Multi-hop relays and shorter feedback loops
• Parallel origin fetches, or racing
• The shared impact of a slow fetch at an edge

Chapters

• 00:00 — Cache misses belong in CDN evaluation
• 01:55 — The origin-to-edge bottleneck for live delivery
• 02:49 — Placing CDN ingest close to the origin
• 04:27 — Multi-hop relays and transport feedback
• 05:46 — How one origin fetch affects an entire edge

Next Steps

• When comparing CDNs, include origin-to-edge fetch performance alongside cached delivery, especially for live video with little margin for delay
• Discuss with your CDN provider whether ingest can move closer to your origin through co-location, adjacent infrastructure or connectivity across cloud VPCs