Cloudflare Workers Adds On-Demand CPU and Memory Profiling

Cloudflare announced on-demand CPU and memory profiling for Workers and Durable Objects on October 9, 2026. The feature reveals function-level resource usage in running applications.

Flamegraphs visualize call stacks and resource consumption. Wider sections indicate functions associated with greater CPU time or memory allocation, helping developers locate bottlenecks.

Production profiling matters because local development environments often differ from real traffic patterns. Developers can initiate profiles through the Cloudflare dashboard or CLI.

In the dashboard, the Observability interface provides options to select a Worker version, profiling type and capture duration. Results can be inspected or downloaded.

Profiling relies on a live execution environment, so low-traffic Workers may not provide sufficient samples. It does not automatically launch a new isolate solely for measurement.

Memory profiling focuses on allocations during the capture window. Problems limited to startup may be missed. Cloudflare has described continuous profiling as a future direction.

Cloudflare published internal examples of using profiles to identify unnecessary work and memory overhead. Any optimization should be validated against latency, errors and resource use after deployment.

Teams can improve reliability through a repeatable cycle of measurement, hypothesis, code change and verification rather than optimizing by intuition alone.