Achieving Scale to Zero and Back Again with Cloud Foundry
summary
Experimental project scale2zero implements true zero-scale on Cloud Foundry by watching HTTP start/stop events from GoRouter/loggregator and registering routes via a dormant proxy that hijacks requests to wake stopped app containers. Demo stops apps after ~30s idle, first wake incurs ~1.88s extra latency, subsequent requests are fast. Planned integration into app‑autoscaler uses a metric‑forwarder and credentials so apps can push custom metrics; in a demo a worker app auto‑scales +2 when a queue metric exceeds 100.
AI-written summary · about
chapters (10)
- 0:00Introduction to Scale to Zero
- 1:25What is Scale to Zero in Cloud Foundry
- 4:14Existing Approaches for Scale to Zero
- 6:42Implementation Details of Scale to Zero
- 8:51Architecture of Autoscaler and Scale to Zero
- 9:56Live Demo of Scale to Zero
- 14:30Custom Metrics Feature Overview
- 19:48Custom Metrics Demo with Queue Length
- 25:37Stratos UI Integration for Autoscaler
- 27:04Future Enhancements and Summary
related talks
43mHow Platforms Work
08mIntroduction to the Cloud Foundry at Scale Track - Julian Fischer, anynines & Matthias Steiner, SAP
01h 29mThinking Outside the Container: Running Standalone Apps on Cloud Foundry
19mThe Taming of the Queue
20mAdvancing the Future of Cloud Foundry
01h 25mWhat's New in CloudFoundry
29mCustom App Autoscaler Using Cloud Controller API - Christopher Decelles
33mCloud Foundry Weekly: Live at CF Day Germany