Benchmark Evaluation of Resource Efficiency with Container Orchestration Policies in Edge Computing Clusters

Abstract

The proliferation of Internet of Things devices and the increasing demand for ultra-low latency applications have driven the paradigm shift from centralized cloud computing to distributed edge computing. In this context, containerization technologies have emerged as the standard for packaging and deploying microservices across heterogeneous edge environments. However, traditional container orchestration systems, originally designed for resource-abundant cloud datacenters, often exhibit significant inefficiencies when deployed on resource-constrained edge nodes. This paper presents a comprehensive benchmark evaluation of various container orchestration policies to assess their impact on resource efficiency in edge computing clusters. By systematically analyzing different scheduling algorithms, including default spread policies, bin-packing strategies, and custom latency-aware algorithms, this study identifies the critical bottlenecks in current orchestration frameworks. Extensive experiments are conducted on a heterogeneous edge cluster testbed, evaluating metrics such as processor utilization, memory overhead, network bandwidth consumption, and application deployment latency. The findings reveal that default cloud-native orchestration policies incur substantial overhead and lead to suboptimal resource allocation at the edge. In contrast, topology-aware and resource-constrained scheduling policies can improve cluster efficiency and reduce deployment latency. This research provides valuable insights for the design of next-generation, edge-native orchestration frameworks tailored for distributed and heterogeneous environments.