crun vs runc: Memory and Performance Benchmarks 2026

💡 Quick Answer: Use crun for faster container startup (50% faster cold start) and lower memory usage (10x less RSS). Use runc for maximum compatibility and when running on cgroup v1. Both are OCI-compliant — switching is a one-line config change in containerd or CRI-O.

The Problem

Teams assume all container runtimes are equal, but crun (written in C) starts containers 50% faster and uses 10x less memory than runc (written in Go). For Kubernetes nodes running 100+ pods, the difference in node overhead is significant.

The Solution

Performance Comparison

Configure containerd to use crun

# /etc/containerd/config.toml
[plugins."io.containerd.grpc.v1.cri".containerd.runtimes.crun]
  runtime_type = "io.containerd.runc.v2"
  [plugins."io.containerd.grpc.v1.cri".containerd.runtimes.crun.options]
    BinaryName = "/usr/bin/crun"

Configure CRI-O to use crun

# /etc/crio/crio.conf.d/01-crun.conf
[crio.runtime]
default_runtime = "crun"

[crio.runtime.runtimes.crun]
runtime_path = "/usr/bin/crun"
runtime_type = "oci"

Install crun

# Fedora / RHEL 9+
sudo dnf install crun

# Ubuntu / Debian
sudo apt install crun

# Verify
crun --version
# crun version 1.14
# spec: 1.0.0
# +SYSTEMD +SELINUX +CAP +SECCOMP +EBPF +CRIU +LIBKRUN +WASM:wasmedge

Kubernetes RuntimeClass

apiVersion: node.k8s.io/v1
kind: RuntimeClass
metadata:
  name: crun
handler: crun
---
apiVersion: v1
kind: Pod
metadata:
  name: fast-start
spec:
  runtimeClassName: crun
  containers:
    - name: app
      image: registry.example.com/app:v1

graph TD
    subgraph runc - Go Runtime
        R_START[Container Start<br/>~100ms] --> R_MEM[Memory<br/>~10MB RSS per container]
        R_MEM --> R_COMPAT[Maximum compatibility<br/>Default everywhere]
    end
    subgraph crun - C Runtime
        C_START[Container Start<br/>~50ms ⚡] --> C_MEM[Memory<br/>~1MB RSS per container]
        C_MEM --> C_FEAT[WASM support<br/>cgroup v2 native]
    end

Common Issues

crun not found after install: Verify path: which crun. containerd needs the full path in config. Restart containerd after config changes.

Pods failing with crun: Check crun supports your seccomp profile. Some older seccomp profiles reference syscalls not handled by crun.

Best Practices

• crun for performance — 50% faster starts, 10x less memory
• runc for compatibility — default everywhere, battle-tested
• Both OCI-compliant — same container images, same behavior
• RuntimeClass for mixed — run both on the same cluster
• RHEL 9+ defaults to crun — Red Hat already made this choice

Key Takeaways

• crun is 50% faster and 10x lighter than runc — written in C vs Go
• Both are fully OCI-compliant — switching requires only a config change
• crun is default on RHEL 9+, Fedora, and Podman
• Use RuntimeClass to run both runtimes on the same Kubernetes cluster
• crun adds WASM support via wasmedge — runc doesn’t support WASM