Reckon-ing Kubernetes at the Edge using Emulated Clusters

Kubernetes is the industry standard cluster orchestrator and it (or its derivatives) have been proposed for orchestrating edge network resources as used in, e.g., IOT deployments. The limited resources and unreliable connectivity in edge environments present challenges to orchestrators including K8s. Their impact is typically studied through emulation on cloud resources but these are difficult to control and highly variable in performance. We present Reckon-K8s, a single-host container-based emulator of Kubernetes clusters over a configurable virtual network. We validate its design and implementation via calibration against a (small) physical cluster, and use it to demonstrate how cluster topology and network conditions influence performance in edge scenarios.