Cloudflare
2026-07-31
Cloudflare Migrates cdnjs to Workers Platform, Handling 9 Billion Requests Daily
Summary
Key Takeaways
Cloudflare has fully migrated cdnjs, one of the busiest open-source CDNs, to its developer platform. cdnjs serves ~12% of all websites, holds 48.3% of the JavaScript CDM market, handles 108,000 requests per second (9 billion daily), spans 330+ Cloudflare data centers, and achieves a 98.6% cache hit rate.
The new architecture runs entirely on Cloudflare's developer platform: R2 as the single source of truth with no size limits and S3 API access; KV for metadata only (package info, version lists, SRI hashes); Workers Cache as a tiered cache in front of Workers; and Cloudflare Workflows for data ingestion, with cron jobs checking npm/GitHub every 10 minutes and persistent execution for failure recovery.
Platform limitations encountered: 1,000 subrequests per Worker and 1,024 steps per Workflow. Cloudflare raised these to 10 million subrequests (paid plans) and 10,000 default steps (configurable to 25,000). LLMs like ChatGPT, Claude, and Cursor heavily use cdnjs for HTML demos, highlighting its role in AI workloads.
Why It Matters
Cloudflare's migration of cdnjs is a defensive move against competitors like Fastly Compute@Edge and AWS Lambda@Edge, using a flagship case to lock developers into its platform. The architecture creates vendor lock-in by storing content in R2 and metadata in KV, making migration to other object stores costly due to S3 compatibility nuances. Workers Cache is non-customizable, limiting flexibility. The blog downplays limitations: subrequest and Workflow step limits are only lifted on paid plans, and tail latency under cache misses is ignored. For LLM-driven workloads, any cache miss latency directly impacts AI response times. Workflows' persistent execution adds cost and time overhead, making large-scale ingestion expensive.
PRO Decision
【Vendors】Competitors like Fastly and Akamai should highlight their open and customizable edge platforms, pointing out Cloudflare's Worker subrequest and Workflow step limits, and R2's S3 compatibility nuances. They can offer reference architectures for similar use cases with standard object storage and flexible caching policies. 【Enterprises】CIOs and architects must conduct zero-trust audits: assess migration costs from R2 to standard S3, test Worker tail latency under peak loads, evaluate Workflows pricing for large-scale ingestion, and adopt multi-cloud strategies to avoid lock-in. For AI apps relying on cdnjs, establish baseline performance for cache misses. 【Investors】Recognize this as a strategy to boost ARPU by pushing paid plans. While showcasing platform capability, note that limitations are lifted only on paid tiers. Monitor whether developer platform revenue growth comes from existing user upgrades or new acquisitions. Increased competition in edge computing may erode Cloudflare's differentiation.
Get 3-5 key AI infrastructure signals weekly →
💬 Comments (0)