Considering the resource-hungry and capability-constraint of Internet of Things (IoT) nodes, their functionalities, which are encapsulated as containerized IoT services, are composed to satisfy user requests. IoT nodes are usually duty-cycled and energy-awareness. Therefore, IoT service allocation to respective IoT nodes should be re-calibrated on-demand through migrating certain IoT services from their hosted IoT nodes to the others, in order to satisfy the functionally diversity of requests. To solve this problem, this paper proposes a Distributed Migration-based Service Allocation (DMSA) mechanism in dynamic IoT networks, where a game-theoretic approach is adopted to achieve the Nash equilibrium of IoT service allocation optimization. Extensive experiments are conducted, and evaluation results demonstrate that our DMSA performs better than the state of art’s techniques in reducing the response latency of requests and improving the resource utilization efficiency.