A MULTI-OBJECTIVE INTELLIGENT LOAD BALANCING ALGORITHM FOR SDN-ENABLED EDGE CLOUD SYSTEMS

Abstract

SDN-based systems with edge-clouds are more commonly used to provide support to latency-sensitive and resource-intensive applications with dynamic and heterogeneous operating conditions. In that case, efficient task assignment must consider concurrently the availability of computational resources, network characteristics, reliability of the service, and the distribution of the workload in the long run. The paper suggests a two-tiered smart scheduling model of SDN enabled edge-clouds. During the first stage, a multi-objective scheduler aided by a controller chooses the node of execution based on shared assessment of CPU usage, memory usage, queue However, there are multiple stages to it: In the first stage, a controller-assisted multi-objective scheduler picks the node of execution by combining the assessment of CPU usage, memory utilization, queue occupancy, delay, bandwidth, reliability, packet degradation, and The second stage will focus on the implementation of machine learning proactive load prediction into the scheduler to enhance its ability to make decisions with future workload variation. The experimental findings indicate that the AI-controlled scheduler is the most performance-friendly in latency and the average response time is 1.4796, and the rates of SLA violation are 0.0497. Of all the adaptive variants, the XGBoost-copiloted model has the best balance-based-behavior with an optimal fairness index of 0.2986 and a low load imbalance factor of 0.0685. These results reveal that reactive SDN-aware scheduling can serve effectively delay-sensitive operation whereas lightweight predictive adaptation can enhance fairness and stable workload over the long term.