Tiered caching organises a CDN into a hierarchy: a viewer hits an edge POP, the edge POP asks a regional parent on cache miss, the regional parent asks origin only if it too is missing. Each tier serves as a cache for the tier below it, multiplying the effective hit ratio that origin sees. Cloudflare named this "Tiered Cache" as a product, but the underlying pattern — shield, fan-in cache, parent POP — appears under different names in every major CDN.

Operationally tiered caching is a topology decision. Cloudflare offers "Smart Tiered Caching" that picks the parent POP automatically based on lowest latency to origin; CloudFront's Origin Shield requires the operator to designate a specific region; Akamai's tiered distribution is configured per content provider. Multi-tier setups (edge → parent → super-parent → origin) reduce origin load further for very high-fan-out events but add latency in the cache-miss path.

The key trade-off is freshness vs efficiency. A deeper tier hierarchy means more places to invalidate when a manifest changes (which happens every segment-duration in live streaming). Most CDNs handle this with versioned URLs (each new segment has a unique URL, so invalidation is unnecessary for segments) and short TTLs on manifests (1–2 seconds). Done well, tiered caching gives 95 %+ origin offload even during live events with millions of concurrent viewers.