I'm looking for a Python-based way to model how application traffic and workload behavior change when Kubernetes settings are adjusted. The simulator should ideally let me vary pod counts, replicas, CPU and memory requests and limits, and similar configuration values, then observe the impact on performance and scaling. Does an existing tool already support this, or would it make more sense to build a custom simulator?
1 Answer
There isn’t one universal tool that covers every Kubernetes configuration and traffic scenario, but several projects can handle parts of the problem. Kube-capacity, Goldilocks, and the Vertical Pod Autoscaler in recommendation mode can help compare resource requests and limits with observed usage. For generating traffic against a real cluster, Locust and Fortio are useful options. For synthetic, discrete-event simulations of autoscaling behavior without running a full cloud environment, look at projects such as Multiverse and autoscaling simulators. You may need to combine a simulator or traffic generator with a dashboard and your own configuration-analysis layer.
The main challenge would be connecting the traffic model, Kubernetes configuration inputs, autoscaling behavior, and visualization into one workflow. That integration may be more work than the individual components.