FAST: AI-based Network Traffic Analysis and Load Balancing Framework Underlying SDN Clusters
D R Shah, Diya Shah, Fenil Ramoliya, Rajesh Gupta, Hetal Brijesh Shah, Sudeep Tanwar, Hossein Shahinzadeh, Deepak Kumar Garg · 2024
In the smart ecosystem environment, critical request handling is essential for seamless network operation. It is necessary that these requests are forwarded efficiently with low latency and high scalability as well as security. Software-Defined Networking (SDN) plays a crucial role in this environment with its dynamic traffic engineering mechanism, Quality of Service (QoS) implementation, Fast-intelligent routing, and, Artificial Intelligence (AI) integration capacity. Designing policies that appropriately prioritize and manage each type of traffic (web, IoT, social media, network) altogether can raise issues such as scalability, security concerns, QoS assurance, policy management, and vendor lock-in. To mitigate these pressing concerns, we propose FAST (Framework for analysing SDN traffic), SDN-clustering based mechanism. In our proposed approach i.e, FAST, Ensemble Learning (EL) implemented mainframe SDN controller works as the central entity for analyzing incoming traffic and diverting it to its respective SDN cluster. In this configuration, each cluster, equipped with limited yet efficient single-domain working SDNs can process a particular type of data and enhance the overall system performance dynamically. The performance evaluation of FAST includes precision, recall, f1-score and accuracy comparison between Machine Learning (ML) and EL algorithms as well as Receiver Operating Characteristic (ROC) curve, Precision-Recall(PR) curve and Confusion matrix.