Total Kubernetes CPU and memory at a glance.

A millicore (the m suffix, as in 250m) is one thousandth of a CPU core, so 1000m equals one vCPU / hyperthread. Kubernetes measures CPU in these fractional units because a Pod rarely needs a whole core — 250m means a quarter of a core of guaranteed share. A plain number like 0.5 is the same as 500m; the calculator normalizes everything to millicores so requests and totals line up.

Mi is a mebibyte — 1024 × 1024 = 1,048,576 bytes — while MB usually means a megabyte of 1,000,000 bytes. Kubernetes memory suffixes Ki / Mi / Gi are binary (powers of 1024); the suffix-free K / M / G are decimal (powers of 1000). They are close but not equal: 1 Gi is about 1.074 GB. Mixing them silently under-provisions, so the calculator keeps the two families distinct and converts both to exact bytes.

A request is what the scheduler reserves for a container — it guarantees that much CPU and memory and is what bin-packing uses to place the Pod on a node. A limit is the hard ceiling: CPU above the limit is throttled, and memory above the limit gets the container OOM-killed. Requests drive capacity planning; limits cap blast radius. The calculator totals both so you can size a node pool against requests and watch the ceiling against limits.

It is invalid — the Kubernetes API server rejects a container whose limit is lower than its request, because you cannot cap a resource below the amount you have guaranteed. The calculator flags any row where the normalized limit falls under the normalized request so you catch the mistake here instead of in a failed apply or an admission webhook rejection.

No. Every quantity is parsed, normalized and totalled 100% client-side in your tab — there is no server, no account and no logging. You can safely paste resource blocks from internal manifests; nothing is uploaded.

A pod's total request and limit is the sum of all its containers' individual values. Multiply that per-pod total by the replica count to get what the whole deployment reserves. For example, 3 replicas at 250m CPU and 256Mi memory each total 750m CPU and 768Mi. Enter the per-container values and replica count into the calculator and it computes the workload totals instantly.

Mi (mebibytes) and Gi (gibibytes) are binary units where 1024Mi equals 1Gi. Divide Mi by 1024 to get Gi, and multiply Gi by 1024 to get Mi. One Gi equals 1,073,741,824 bytes, which is slightly larger than 1G (1,000,000,000 bytes). The calculator normalizes all memory values to exact bytes internally, so Mi and Gi inputs compare and total correctly without any manual conversion.

Yes, the calculator is completely free. It runs entirely in your browser with no server, no account and no data upload — you do not need to sign up or log in. You can safely paste resource values from internal manifests because nothing leaves your device.

Base requests on observed p95 or p99 usage gathered with kubectl top over a representative 24-hour period, then set limits at roughly two to five times the request value. Requests set too high waste schedulable node capacity; set too low they starve the pod under load and cause CPU throttling or OOMKills. Use the calculator to total per-container values across replicas before committing numbers to a manifest.