kbench

kbench is a benchmarking tool for measuring the control plane performance of a Kubernetes cluster.

Installation

kbench is available on PyPI.

$ pip3 install kbench

Usage

pod-throughput

Launch multiple pods in parallel and measure their startup and cleanup time.

$ kbench pod-throughput

• -n, --num-pods: Number of pods to launch.
• -i, --image: Container image to use.
• --timings / --no-timings: Print timing information for all pods.

pod-latency

Launch multiple pods sequentially and measure their startup and cleanup time.

$ kbench pod-latency

• -n, --num-pods: Number of pods to launch.
• -i, --image: Container image to use.
• --timings / --no-timings: Print timing information for all pods.

deployment-scaling

Create a deployment and measure scale-in/out latency. First, a deployment with m replicas is created. Then, the deployment is scaled-out to n replicas. Once the scale-out is completed, the deployment is scaled-in to m replicas again.

$ kbench deployment-scaling

• -i, --image: Container image to use.
• -m, --num-init-replicas: Initial number of replicas.
• -n, --num-target-replicas: Target number of replicas.