From Bottleneck To Breakthrough: Conquering Applications Star...
summary
Explains a queue-based method implemented with Karpenter to handle applications that need high CPU/memory only during startup (examples: Java, .NET, Node.js, large data-processing or ML workloads). The approach lets workloads request typical resources rather than peak startup values by temporarily spreading pods across multiple smaller nodes to avoid single-node contention, then consolidating them onto larger nodes after initialization to reduce wasted capacity.
AI-written summary · about
related talks
33mNavigating and Optimizing Karpenter - Avoiding Common Pitfalls
36mGroupless Autoscaling with Karpenter
38mHow Can Pod Start-up Be Accelerated on Nodes in Large Clusters?
36mAutoscaling Isn't Set-and-Forget: Tuning KEDA & Karpenter for Real Workloads.
36mRise of the Machines [B]
29mThe only easy day was yesterday (Zuul+Karpenter)
18mAutomating Kubernetes Cluster Updates: Achieving Zero Downtime Effortlessly
33mKubernetes cost optimization - How to actually do it