AN ADAPTIVE LOAD BALANCING ALGORITHM FOR HETEROGENEOUS CLOUD ENVIRONMENTS

Abstract

Efficient resource allocation in heterogeneous cloud environments remains challenging due to dynamic workloads, uncertainty, and diverse Quality of Service requirements. Traditional load balancing methods such as Round Robin fail to consider system dynamics and request characteristics, while adaptive and predictive approaches often overlook fairness and semantic context. This paper proposes a Request-Aware Fuzzy Round Robin algorithm that integrates request semantics, fuzzy logic, and predictive load estimation into a fairness-preserving framework. The method incorporates priority, delay sensitivity, and deadline constraints alongside server state information and EWMA-based prediction. Evaluation in CloudSim Plus against RR, Priority-based RR, and Predictive Adaptive Load Balancer shows consistent improvements in response time, waiting time, makespan, throughput, and load balance under heterogeneous and burst workloads.