An optimized bias-resilient double-indexed throttled scheduling algorithm for fair and efficient load balancing in cloud datacenters

As cloud computing environments continue to support increasingly dynamic and diverse workloads, efficient task scheduling and load balancing remain critical for ensuring system responsiveness and scalability. Classical rule-based schedulers such as Throttled and Round Robin provide low overhead but suffer from systematic positional bias and uneven resource utilization under sustained load, while Equally Spread Current Execution (ESCE) incurs high scheduling overhead and excessive resource consumption. This paper proposes the Double-Indexed Throttled Load Balancer (DITLB), a bias-resilient VM allocation strategy that enhances the traditional Throttled policy through bidirectional indexed scanning, a Rotating Start Pointer (RSP) to eliminate index-based bias, and a Last Success Cache (LSC) to exploit temporal locality and reduce redundant VM probing. An analytical model is further introduced to study the effect of cache size on allocation efficiency and fairness. Simulation experiments conducted on CloudSim across light, medium, high and congested workload scenarios, each evaluated over ten independent runs, demonstrate that DITLB consistently achieves more homogeneous load distribution and near-perfect VM-level fairness compared to standard Throttled and Round Robin scheduling, while preserving stable overall response time. Statistical validation using Wilcoxon signed-rank tests and Cliff’s delta confirms that DITLB introduces no response-time penalty while significantly mitigating allocation bias. In contrast, ESCE exhibits substantially higher datacenter cost in small-to-medium workloads. Overall, DITLB provides a cost-neutral, fairness-aware, and scalable scheduling solution for modern cloud and edge-enabled datacenters.