Scaling Kubernetes Workloads with Node Swap
By BIOS Founding Team
The post argues that memory is the first hard limit in many clusters, and that agentic AI workloads make it worse. Node swap on fast local SSDs helps, and it has been GA since v1.34. The config uses failSwapOn: false with memorySwap.swapBehavior: LimitedSwap, which applies to Burstable pods.
Reported benchmarks on a GKE c4-standard-32 node include 50 percent lower RAM for a kernel build and up to 3x pod density for gVisor Python sandboxes. The authors warn that swap is insurance rather than a RAM replacement, and that results are hardware-specific.
Full details: https://kubernetes.io/blog/2026/10/05/scaling-kubernetes-workloads-with-node-swap/
More from the community
Docker Hub Security Advisory: Personal Access Token Transfer
Docker Hub's account-to-organization conversion could reassign a converted account's personal access tokens to a named new owner without rotating the secret or notifying that owner. Docker rated the issue critical.
Vue 3.6.0-rc.11
The latest release candidate adds one feature: Vapor slot content can now transition inside a VDOM Transition. It also has about 60 bug fixes, mainly in the Vapor compiler and runtime.
Python 3.15.0 Is Here
Python 3.15.0 is the stable release of the 3.15 series, built from 5,643 commits by 1,012 contributors. Headline changes are explicit lazy imports (PEP 810), a frozendict built-in (PEP 814), a sentinel built-in (PEP 6...