An improved Lagrangian relaxation algorithm based SDN framework for industrial internet hybrid service flow scheduling

Abstract

Abstract The Industrial Internet is the key for Industry 4.0, and network control in the industrial internet usually requires high reliability and low latency. The industrial internet ubiquitously connects all relevant Internet of things (IoT) sensing and actuating devices, allowing for monitoring and control of multiple industrial systems. Unfortunately, guaranteeing very low end-to-end wait times is particularly challenging because the transmissions must be articulated in time. In the industrial internet, there usually coexist multiple streams. The amount of data for controlling business flows is small, while other business flows (e.g., interactive business flows, sensing business flows) typically transmit large amounts of data across the network. These data flows are mainly processed in traditional switches using a queue-based "store-and-forward" mode for data exchange, consuming much bandwidth and filling up the network buffers. This causes delays in the control flow. In our research, we propose an Software Defined Networking (SDN) framework to reduce such delays and ensure real-time delivery of mixed service flows. The scheduling policy is performed through the northbound Application Programming Interface (API) of the SDN controller so that the dynamic network topology can be satisfied. We use the concept of edge and intermediate switches, where each switch port sends data at a specific time to avoid queuing intermediate switches. We also introduce an improved Langerian relaxation algorithm to select the best path to ensure low latency. Finally, the path rules are deployed to the switches in flow tables through the SDN controller. Our mathematical analysis and simulation evaluation demonstrates that the proposed scheme is more efficient.