Lightning Talk: Predictive Autoscaling in Kubernetes With KEDA and Prophet
summary
Combines Prometheus metrics, a lightweight Python cron job that runs a Prophet forecast, and an adapter that exposes predicted values to KEDA which feeds Kubernetes HPA so replicas adjust before load spikes; forecasts run 15 minutes ahead and the pipeline uses a Python service + scaledObject external scaler to proactively change allocation; claims reduced latency and better resource use, demonstrates KEDA reading external forecasts from Prophet and works across clusters with KEDA integration and HPA cooldown and polling parameters.
AI-written summary · about
related talks
07mScaling Predictive Manufacturing Analytics: Kubernetes-Powered CRM
09mLightning Talk: Optimize the Ray Schedule and Autoscaling on Kubernetes Cloud: A Hetero...
31mFailing to Autoscale?
39mProphet at scale to tune & forecast time series at Spotify
38mBeyond Kubernetes: Adapting To Spe...
35mOptimizing Metrics Collection & Serving When Autoscaling LLM Workloads
26mDynamically Tuning Pods: Leveraging Time Series ML Models with KubeFlow
31mShared GPU Scheduling & Proactive Autoscaling: A Production Blueprint for 1000+…