The edge-based computing paradigm (ECP) becomes one of the most innovative modes of processing distributed Interneit of Things (IoT) sensor data. However, the edge nodes in ECP are usually resource-constrained. When more services are executed on an edge node, the resources required by these services may exceed the edge node’s, so as to fail to maintain the normal running of the edge node. In order to solve this problem, this paper proposes a resource-constrained smart service migration framework for edge computing environment in IoT (IoT-RECSM) and a dynamic edge service migration algorithm. Based on this algorithm, the framework can dynamically migrate services of resource-critical edge nodes to resource-rich nodes. In the framework, four abstract models are presented to quantificationally evaluate the resource usage of edge nodes and the resource consumption of edge service in real-time. Finally, an edge smart services migration prototype system is implemented to simulate the edge service migration in IoT environment. Based on the system, an IoT case including 10 edge nodes is simulated to evaluate the proposed approach. According to the experiment results, service migration among edge nodes not only maintains the stability of service execution on edge nodes, but also reduces the sensor data traffic between edge nodes and cloud center.
Adaptive Federated Learning in Resource Constrained Edge Computing Systems
