What is the difference between requests and limits?

Requests are what the scheduler uses to place the pod on a node — the pod is guaranteed this amount. Limits cap the maximum the container can consume. A container that exceeds its memory limit is OOMKilled; exceeding CPU limits causes throttling, not termination.

How much headroom should I add?

A 25–50% buffer over observed peak is typical for stateless services. Latency-sensitive APIs often need more (50–100%) to handle traffic spikes without CPU throttling. Batch jobs can run leaner (10–20%) because slower is acceptable.

Should limits be much higher than requests?

A limits/requests ratio of 1.5–2x for CPU and 1.25–1.5x for memory is common. Very high ratios (e.g. 10x) mean the node may be overcommitted, risking eviction under pressure. Very low ratios (1.0x) risk unnecessary throttling on legitimate spikes.

What units does Kubernetes use?

CPU is specified in millicores (m): 1000m = 1 vCPU core. Memory is in bytes with suffixes: Mi (mebibytes), Gi (gibibytes). This calculator outputs in those standard units.

Why does my web server need more CPU than a batch job with the same concurrency?

Web servers often have higher per-request overhead from TLS, HTTP parsing, and connection handling, while batch jobs amortize overhead over longer tasks. CPU sizing is calculated directly from your observed peak CPU and headroom, while application type provides workload-specific guidance and memory heuristics.

Kubernetes Resource Sizing Calculator

Estimate CPU and memory requests/limits for Kubernetes pods based on app type, concurrency level, and observed usage.

Workload Parameters

Application Type

Peak Concurrency (simultaneous requests/tasks)

Observed Peak CPU (millicores per replica)

From your monitoring (e.g. Prometheus container_cpu_usage_seconds_total). 1000m = 1 vCPU.

Observed Peak Memory (MiB per replica)

From your monitoring. Enter 0 if unknown — a heuristic will be used.

Safety Headroom (50%)

10%100%

Limits / Requests Ratio (1.5×)

1.0×4.0×

Fill in the parameters and click Calculate Resources.

Copied!

Summary

Kubernetes resource sizing determines the CPU and memory requests and limits you declare in a pod spec. Getting these numbers right prevents OOMKill events, avoids CPU throttling, and ensures the scheduler places pods on nodes with enough capacity. This calculator estimates those values from your application type, concurrency, and observed peak usage — with a configurable safety buffer on top.

How it works

Select your application type (web API, worker, batch job, etc.).
Enter the expected peak concurrency (requests or tasks in flight simultaneously).
Input your observed peak CPU and memory usage from monitoring.
Choose a safety headroom percentage to buffer for spikes.
The calculator outputs recommended requests and limits for your pod spec.

Use cases

Right-size pods before deploying to production to avoid OOMKill events.
Convert raw Prometheus/Datadog metrics into pod resource fields.
Estimate cluster capacity needed for a given workload count.
Tune the requests/limits ratio to improve bin-packing on nodes.
Prepare resource specs for HorizontalPodAutoscaler threshold planning.
Compare resource profiles across different application types.

Frequently Asked Questions

Last updated: 2026-06-15 · Reviewed by Nham Vu