A PREDICTIVE LOAD BALANCING STRATEGY FOR SUSTAINABLE SOFTWARE-DEFINED NETWORKS

Abstract

The relentless expansion of cloud computing is pushing digital infrastructure to its limits, making advanced server load balancing essential for speed and overall system sustainability. Conventional methods operate reactively, including Least Response Time and Least Connection algorithms. Their failure to anticipate sudden traffic spikes often leads to performance degradation, a problem magnified in environments with mixed server capabilities. Confronting this challenge head-on, our research proposes a novel Predicted Load Balancing (PLB) algorithm. PLB is a proactive solution that uniquely fuses live network data with predictive analytics to distribute traffic intelligently before bottlenecks occur. We rigorously evaluated PLB against traditional algorithms using Mininet to emulate a Software-Defined Networking (SDN) architecture. Tests across both uniform and heterogeneous server setups yielded compelling results. PLB consistently outperformed its counterparts, slashing response times by 12-20% and boosting overall throughput by 8-15%. Its most notable achievement was a staggering 60% improvement in throughput within complex, heterogeneous environments. Furthermore, PLB achieved a 30% reduction in server load variance, a key metric that directly translates to greater stability and fewer errors. These findings confirm that PLB fundamentally shifts the load balancing paradigm from a reactive to a predictive model. This capability is transformative, paving the way for constructing more resilient and energy-efficient networks. By preemptively managing traffic, PLB offers a robust framework for sustainable server management that maintains reliable performance even under the most dynamic and demanding conditions.