tag
5 posts
Plan Linux HugeTLB pools by page size and NUMA node, distinguish reservations from faults, and validate hugetlb cgroup limits before deployment.
Configure Kubernetes static CPU management and NUMA alignment, understand Guaranteed Pod eligibility, and prevent topology admission failures in production.
Use FreeBSD cpuset masks and NUMA domain policies deliberately, verify effective placement, and measure whether affinity improves real workloads.
Operate Linux memory hotplug safely by inspecting block granularity, online policy, NUMA placement, movable-page limits, and rollback conditions first.
Diagnose NUMA locality by separating where threads run from where pages are allocated, then measure policy, cpuset constraints, and remote access.