Cost- and Latency-Aware Scheduling Plugin for Microservice Applications in Edge Computing on K3s
Amira Rayane Benamer · 2025
Edge computing has emerged as a key solution for IoT applications, particularly in addressing stringent Quality of Service (QoS) requirements, such as low-latency demands for real-time decision-making. By processing data closer to users, edge computing minimizes response times while efficiently managing the vast and continuous influx of data from IoT devices. To support scalable, resilient, and cloud-native architectures, microservice-based applications (MSA) have gained traction in IoT development. However, deploying these applications in resource-constrained edge environments requires efficient orchestration to optimize resource utilization while meeting QoS constraints. In this work, we adopt K3s, a lightweight Kubernetes (K8s) distribution designed for edge environments, and identify limitations in its default scheduling mechanism. Specifically, existing schedulers overlook latency and resource usage costs when placing microservices across heterogeneous nodes. To address this gap, we develop a customized scheduling plugin that enhances microservice placement by prioritizing both low-latency execution and cost efficiency. Our experimental evaluation demonstrates that the proposed scheduler outperforms baseline solutions, including state-of-theart alternatives and the default K3s scheduler, in terms of response time and resource efficiency. This work serves as a foundation for dynamic and adaptive scheduling strategies in future edge computing environments.