Nowadays, mobile device manufacturers employ workflow scheduling. This involves moving data from devices to cloud centers using fog nodes. Such a system facilitates a serverless workflow scheduling scheme. However, existing scheduling schemes have not yet considered the communication problems between cloud and fog nodes, and neglected the heterogeneity of node resource(e.g., CPU, memory, and bandwidth) under limited conditions. To solve these problems, we propose a novel low migration and low energy consumption framework aiming at improving the performance of meta-heuristic algorithms for workflow scheduling in fog computing. Under resource-constrained conditions, we focus on maximizing the task completion efficiency and improving the resource utilisation of the heterogeneous node resource, and consider the communication link latency between the cloud and fog nodes. As a preprocessing part, we propose a “Memory segmentation algorithm” that partitions workflow tasks based on their memory usage patterns. This segmentation strategy reduces the task migration time, and both improves the task allocation accuracy. In addition, we use the “Heterogeneous node resource service score evaluation model” to quantify the service quality and accurately evaluate the scheduling strategy to help the algorithm converge. Comparative and ablation experiments on the publicly available “Bitbrains” dataset suggest that our framework can help the original naive algorithm increase task completion by 50%, reduce task migration times by 27% and reduce energy loss by 12%.