Smart Autoscaling for Kubernetes: From Reactive to Predictive Scaling

Join Zbynek Roubalik, founder and CTO at Kedify, for a technical session on building intelligent, efficient autoscaling for Kubernetes. Learn how to move beyond traditional CPU and memory scaling to smarter, proactive strategies that adapt to real-world traffic and events.

This session combines practical demos and real-world examples to show how Kedify, built on top of KEDA, can significantly reduce cloud costs while maintaining top performance.

You will learn how to:

• Implement custom metrics, event-driven and HTTP-based autoscaling, and seamless scale-to-zero without losing traffic.
• Understand why CPU and memory-based scaling is reactive and often inefficient.
• Scale any type of workload, from REST APIs and microservices to event-driven systems, LLM inference, and GPU-intensive workloads.
• Apply predictive autoscaling with time-series forecasting to scale before load spikes occur.
• Explore multi-cluster scaling and emerging trends in cloud cost optimization.

You can also start building now with the Kedify HTTP Autoscaling learning path—the same workflow we’ll cover in the recording." with "Watch the on-demand session below, or start building with the Kedify HTTP Autoscaling learning path and follow the same workflow at your own pace.