Kedify Achieves SOC 2 Type II Certification!
Learn more about our commitment to security
All Events
From Reactive to Proactive: Forecast-Driven Scaling for Kubernetes
Join Christian Melendez (Principal Specialist Solutions Architect at Amazon Web Services) and Jan Wozniak from Kedify at ContainerDays London 2026 for a technical session on advancing autoscaling paradigms.
The premise: reactive autoscaling is a lagging indicator
Most teams start with the “classic” approach:
- scale pods based on CPU/memory
- then rely on cluster autoscaling to add nodes when pods don’t fit
That workflow is inherently reactive. Even when it’s configured correctly, it can still be late because the signal arrives late, the decision happens late, and the capacity shows up late.
In our talk, we’ll break down where that latency actually comes from:
- metric delays (collection + aggregation + evaluation)
- image pulls and cold caches
- node startup + bootstrap time
- scheduling constraints that “hide” real demand until it’s too late
Where we’re going: day-2 excellence for Karpenter + KEDA
We will go beyond basic setup and into the practical patterns that make autoscaling stable in production: troubleshooting, cost control, efficiency, and (eventually) forecasting.
