Web performance optimization is a constant challenge for companies relying on public cloud. With global traffic growth, caching strategies have become essential to reduce latency and infrastructure costs. In this context, the Smart Tiered Cache for Public Cloud Regions feature, developed by Cloudflare, represents a significant advancement. However, for organizations seeking to maximize this technology, having the support of a technology partner like Q2BSTUDIO can make a difference. Our experience in cloud services AWS and Azure allows us to design architectures that efficiently integrate these solutions.
The main problem this innovation addresses is the difficulty of applying tiered caching when origins are behind anycast or regional unicast IP addresses, as is the case with load balancers from major cloud providers (AWS, GCP, Azure, Oracle). In these cases, Cloudflare’s latency probes detect multiple close locations simultaneously, preventing the selection of a single optimal data center as the upper tier. This causes so-called “hairpinning”: traffic takes unnecessary round trips across continents, adding hundreds of milliseconds of latency. For example, a request from Asia heading to an origin in Singapore may end up passing through Chicago, doubling the distance traveled.
The presented solution allows administrators to manually indicate the cloud region where the origin resides, by adding a “region hint.” With that information, Cloudflare can correctly map the anycast IP to its physical location and select the closest upper tier, plus a fallback tier at a different point of presence. This improves cache hit ratios, reduces connections to the origin, and eliminates hairpinning. The system automatically updates with the IP ranges of each cloud provider and combines continuous probing data with a weighted voting algorithm to choose the best tier.
For companies developing cloud applications, this functionality is key to delivering fast and reliable user experiences. At Q2BSTUDIO, we implement custom software that integrates these optimization mechanisms, combining them with artificial intelligence and AI agents to predict traffic patterns and dynamically adjust cache configuration. Additionally, our cybersecurity and Business Intelligence (Power BI) solutions directly benefit from lower latency and greater data access efficiency.
From a technical standpoint, implementation is straightforward: in the Cloudflare dashboard, go to Caching > Tiered Cache > Origin Configuration, identify the anycast IP, and assign the corresponding cloud region (e.g., aws:us-east-1). Configuration can be done per IP or in bulk, and is also available via API and Terraform, facilitating integration into infrastructure-as-code workflows. Cloudflare refreshes IP ranges every few hours and latency probes every 15 minutes, ensuring the upper tier selection remains optimal even when cloud providers add or remove subnets.
For regions that do not yet have enough probing data, the system falls back to a geographic selection based on the nearest Tier 1 PoPs. As the origin begins to receive traffic and metrics accumulate, the region automatically switches to data-driven selection. This hybrid approach ensures no performance loss during the startup period.
Beyond the technical aspects, this feature has a direct business impact: lower latency means better conversion rates, reduced origin server load, and lower bandwidth costs. Companies using cloud services like AWS or Azure can particularly benefit if they have a global presence and need to serve static or dynamic content from distributed regions.
At Q2BSTUDIO, we help our clients implement these optimizations as part of a comprehensive digital transformation strategy. Our team combines deep knowledge of cloud platforms, custom software development, artificial intelligence, and cybersecurity to deliver robust and scalable solutions. If your company relies on public cloud origins and seeks to improve cache performance, we invite you to explore how Smart Tiered Cache can be integrated into your architecture with expert support.
