# Architecting on Cloudflare

> An independent book on Cloudflare's Developer Platform covering architectural decision frameworks, trade-off analysis, and migration playbooks from AWS, Azure, and GCP.

Published: 2026-06-08
URL: https://daniliants.com/insights/architecting-on-cloudflare/
Tags: cloudflare, edge-computing, serverless, cloud-architecture, workers, platform-evaluation, decision-frameworks, devops

---

## Summary

An independent book by a solution architect at a Cloudflare partner covering the Cloudflare Developer Platform: Workers, Durable Objects, D1, R2, Queues, Workflows, Containers, and the AI stack. Explicitly not documentation: focuses on architectural decision frameworks, trade-off analysis, and migration playbooks from AWS/Azure/GCP. Reflects the platform state as of May 2026 including the Agents Week announcements.

## Key Insight

- **Core heuristic**: "start with Cloudflare and find reasons not to use it" rather than defaulting to AWS and evaluating migration later. The logic: Cloudflare's defaults (global deployment, instant scaling, zero cold starts, integrated security) are costly to replicate on hyperscalers, so switching costs accumulate before you even evaluate alternatives.
- **Scope distinction matters**: the book covers only the Developer Platform (compute, storage, AI), NOT the security/networking stack (WAF, DDoS, Zero Trust). Two very different product categories under the same brand.
- **V8 isolate model** is the foundational architectural difference from Lambda. Sub-millisecond cold starts come from the isolate model, not marketing. The CPU-time billing model (vs wall-clock on Lambda) has direct economic consequences.
- **Durable Objects** are called out as genuinely novel coordination primitives with no equivalent elsewhere, relevant for stateful edge scenarios like real-time collaboration, presence, rate limiting.
- **D1 (SQLite at edge)** is positioned as potentially superior to PostgreSQL for certain read-heavy, globally distributed workloads. R2's zero egress fees change cost calculations vs S3 at meaningful data volumes.
- **AI stack on Cloudflare** (Workers AI + Vectorize + Agents SDK) is framed as lower-infrastructure overhead than running your own RAG pipeline, but the book explicitly acknowledges it may not be sufficient when GPT-4 or Claude-class models are needed.
- **Honest limitation catalog** lives in Chapter 24, where the author committed to listing hard thresholds where the platform stops being the right choice. The author's commercial relationship with Cloudflare raises the stakes for honesty, not lowers them.
- **Platform currency**: reflects May 2026 including R2 SQL multi-table joins, AI Gateway REST API, Dynamic Workflows.