In this paper, we propose a scheme to address the problem of load management in the control plane of Software-Defined Internet of Things (SDIoT) networks. In SDIoT, multiple controllers are deployed to enhance network scalability. With the growth of IoT, the number of devices is increasing rapidly. The management of control plane load is an essential issue for IoT networks because of the dynamic traffic characteristics. IoT traffic is highly dynamic due to the heterogeneity of IoT devices in terms of mobility, activation model, Quality of Service (QoS) demand, and flow generation rate. The challenge is to prevent controller overload and distribute traffic optimally under the consideration of heterogeneous IoT devices. The proposed scheme estimates control plane load based on the mobility and activation model of IoT devices. For mobility prediction, we use Order-m fallback Markov Predictor as it consumes less space and performs efficiently even for small values of m. Based on the prediction results, we implement a traffic-aware rule-caching mechanism and a master controller assignment scheme to reduce the control plane load. Simulation results show that the proposed scheme reduces the peak intensity of the control traffic by 23.08% and 16.67%, as compared to the considered benchmark schemes.

中文翻译：

---

核心：软件定义物联网网络中基于预测的控制平面负载减少

在本文中，我们提出了一种解决软件定义物联网 (SDIoT) 网络控制平面中的负载管理问题的方案。在 SDIoT 中，部署多个控制器以增强网络可扩展性。随着物联网的发展，设备的数量正在迅速增加。由于动态流量特性，控制平面负载的管理是物联网网络的基本问题。由于物联网设备在移动性、激活模型、服务质量 (QoS) 需求和流量生成率方面的异构性，物联网流量是高度动态的。面临的挑战是在考虑异构物联网设备的情况下，防止控制器过载并优化分配流量。所提出的方案基于物联网设备的移动性和激活模型来估计控制平面负载。对于流动性预测，我们使用 Order-m 回退马尔可夫预测器，因为它消耗更少的空间并且即使对于 m 的小值也能高效执行。基于预测结果，我们实现了流量感知规则缓存机制和主控制器分配方案，以减少控制平面负载。仿真结果表明，与所考虑的基准方案相比，所提出的方案将控制流量的峰值强度降低了 23.08% 和 16.67%。