
Cloudflare adds public cloud region hints to Smart Tiered Cache
Cloudflare Smart Tiered Cache now uses public cloud region hints to improve cache routing for AWS, Google Cloud, Azure and Oracle origins.
Cloudflare has updated Smart Tiered Cache so customers can steer cache routing for origins hosted behind public-cloud anycast or regional unicast networks. The company said on July 10 that Smart Tiered Cache for Public Cloud Regions is available now for origins on AWS, Google Cloud, Microsoft Azure, and Oracle Cloud.
The change targets a specific problem in content delivery: a cloud origin's IP address may not point cleanly to the real place where an application is running. Cloudflare's existing Smart Tiered Cache chooses one upper-tier data center by probing latency to the origin IP. That works well for fixed unicast origins, but public cloud front ends can make the same IP appear close to many Cloudflare locations at the same time.
Why it matters
When the origin location is ambiguous, Cloudflare said its cache system has had to fall back to multiple upper tiers. That keeps delivery working, but it can lower cache efficiency and send more requests back to the origin. In a worst-case example described by the company, an origin running in Singapore could be selected through an upper tier in Chicago, causing traffic from Asian users to cross oceans unnecessarily before reaching the backend.
The new feature lets operators provide a cloud region hint, such as an AWS or Google Cloud region identifier. Cloudflare then maps that region to primary and fallback upper tiers rather than relying only on latency measurements to a potentially anycasted IP. The company says hints can be set in the dashboard under Caching and Tiered Cache, one IP at a time or in bulk, and are also exposed through its API and Terraform for infrastructure-as-code workflows.
How Cloudflare picks the route
Cloudflare says it fetches IP range files from supported cloud providers every few hours and combines that data with an upper-tier database refreshed by continuous latency probing every 15 minutes. Each matching cloud subnet contributes to the choice of a region's primary upper tier, while primary and fallback selections are kept in different points of presence to avoid a single local failure taking out both paths.
For regions without enough probe data, Cloudflare falls back to the closest Tier 1 point of presence by geography, then shifts to measured data once more customer origins are visible. The announcement is a practical CDN improvement rather than a flashy product launch, but it matters for teams trying to reduce origin load and latency as more applications sit behind public-cloud front doors.
Sources
Cover photo by panumas nikhomkhai on Pexels, used under the Pexels License.
CyberOGZ Team






Comments (0)
Leave a Comment