A Controllable and Realistic Framework for Evaluating Microservice Scheduling in Cloud-Edge Continuum

The transition from traditional architectures to containerized microservices within the cloud-edge computing continuum introduces significant challenges, particularly in the efficient scheduling of microservices under dynamic conditions. Complex and fluctuating call-graph dependencies, varying cross-node communication latencies, and unpredictable bandwidth conditions substantially impact the performance and reliability of deployed microservices. Consequently, accurately evaluating scheduling policies in such dynamic environments remains essential yet challenging due to the lack of realistic and controllable evaluation frameworks. In this paper, we propose iDynamics, a novel evaluation framework designed explicitly to address these challenges. iDynamics provides comprehensive and controllable evaluation capabilities by emulating realistic dynamics, including configurable call-graph topologies, cross-node communication delays, and bandwidth variability. The framework is composed of modular components, such as the Graph Dynamics Analyzer, Networking Dynamics Manager, and Scheduling Policy Extender, enabling fine-grained environmental control and facilitating systematic comparisons of different scheduling strategies. Extensive experiments on a real cloud-edge testbed demonstrate that iDynamics effectively captures diverse dynamic scenarios encountered in microservice deployments, offering a robust solution for evaluating and optimizing policy performance under realistic and controllable conditions.