Optimizing Transmission for The IoT Using P4 Switches and Federated Averaging

Optimizing Transmission for The IoT Using P4 Switches and Federated Averaging

In the 5G era, software virtualization and machine learning, are becoming more mature. Ah Software-Defined Network, allows network operators to allocate network functions and adjust topology flexibility more flexibly. Machine learning classifies different learning styles and learning tasks using different types of framework, so machine has many different applications, and a relatively short time is required to analyze data. However, Big Data creates other problems. In terms of machine learning, large amounts of data, increase convergence time so questions about private data are not solved, and models must perform more effectively.
This study uses P4 (programming protocol-independent packet processors) switches to develop protocols to make effective and flexible changes to the action rules for packets. Federated averaging is also used to maintain the privacy of the terminal and for model training to address the problems of a centralized and decentralized learning framework. This study also uses P4 and federated learning to design an environment for the optimization of packet transmission for the Internet of Things. The experimental results, show that these two technologies improve the efficiency with which packets are transmitted for the Internet of Things.