Container Resource Sizing: CPU & Memory Requests, Node Packing & Cost
Millicores and Mi/Gi, requests vs limits, p95 right-sizing, pods per node, node counts with DaemonSets and N+1, first-fit bin packing, cost per service and autoscaler replicas through the day: 8 graded labs on one fictional cluster.
The first lesson is free — no account needed.
About this course
A container resource sizing course that right-sizes one fictional company's Kubernetes cluster in a spreadsheet, for platform and DevOps engineers, SREs, FinOps analysts and cloud students. You convert CPU and memory requests written as millicores, Mi, Gi, M and G into cores and MiB, compare requests and limits with measured usage to find idle reserved CPU and overcommit, right-size eight services from p95 usage times headroom rounded up to clean steps, work out allocatable capacity and pods per node and which resource runs out first, calculate the node count after DaemonSets, a target fill and one spare node, place ten pods with first-fit bin packing, split the node bill into a price per vCPU and per GiB for showback and plan autoscaler replicas through a busy day with the HPA formula. Independent course, not affiliated with the CNCF or the Linux Foundation. The first lab is free.
Lessons
- CPU Millicores & Memory Units: m, Mi, Gi, M and GFree
- Requests vs Limits: Utilisation, Idle Reserve & Overcommit Pro
- Right-Sizing Requests: p95 Usage × Headroom, Rounded Up Pro
- Pods per Node: Allocatable Capacity & the Binding Resource Pro
- How Many Nodes? Total Requests, DaemonSets, Target Fill & N+1 Pro
- Bin Packing: First-Fit Placement, Step by Step Pro
- Cost per Service: Allocating Node Cost by Requests Pro
- Autoscaling: the HPA Formula & Replicas Through the Day (Chart) Pro