Part of the Technology & IT suite · 35 calculators

Container Resources Calculator

Total CPU and memory a set of container replicas requests, how many nodes that needs, and what is left over once the system reserves its share.

Each replica requests a share of CPU (measured in millicores, where 1000m is one core) and memory.

1000m = 1 core
Results update as you type
Results
Nodes needed
2
Total CPU requested
Total memory requested
Limited by
Replicas that fit per node
Spare CPU across the cluster
Spare memory across the cluster
Reviewed September 2026. Computing arithmetic: bytes, bits, seconds and probabilities do not change by country. Australian broadband speeds are quoted in megabits per second (Mbps).
No account required · Google Analytics off unless allowedCalculator arithmetic runs in your browserResults update as you type
All calculations run 100% in your browser. The calculator code does not submit your figures to GlobalCalc to obtain a result.
About container resources

How the container resources calculator works

Each replica requests a share of CPU (measured in millicores, where 1000m is one core) and memory. Multiply by the replica count for the workload total. Nodes cannot give all of their capacity to workloads: the operating system, the kubelet and system daemons reserve a portion, typically 10–20%. The node count is then the total request divided by the allocatable remainder, taking whichever of CPU or memory runs out first.

Formula: nodes = ⌈max(total CPU ÷ allocatable CPU, total memory ÷ allocatable memory)⌉

Worked examples

InputsNodes neededNote
20 replicas of 500m / 512 MB on 8-core nodes210 cores requested → 2 nodes
Memory-heavy workload3memory becomes the limit
A single large replica22 nodes

Frequently asked questions

What is a millicore?

A thousandth of a CPU core. 500m is half a core; 2000m is two cores.

Why can I not use the whole node?

The kubelet, container runtime and operating system reserve capacity — typically 10–20%. Requesting the full node size means pods never schedule.

Should I set requests or limits?

Requests decide scheduling and therefore this calculation. Limits cap usage; setting them equal gives predictable performance at the cost of flexibility.

What if CPU and memory are unbalanced?

The "limited by" row tells you which runs out first. Rebalancing requests, or choosing a node type with a different ratio, saves real money.

Does this include headroom for failures?

No — add at least one node so the cluster survives losing one, the same N+1 logic as server capacity.

Where these figures come from

Last checked: September 2026. Units follow the SI decimal convention (IEC 80000-13 defines the binary alternatives); network and security figures cite the defining standard.